Student-built projects, with real partners and real impact.

· Spring 2026

APLL

The goal is to create a digital platform (web-based), that helps students learn and get evaluated on topics required for their university lectures. It ensures that students are ready and prepared for any given lecture, don’t learn anything out of order, and understand all the pre-requisite knowledge. It helps professors ensure that the basic knowledge required for their lecture is understood by the students, and provides a centralized platform for them to organize the grades for their students' pre-lecture assignments.

Binyomin Abrams

Autograding Tool for Written Answers and Complex Assignments

In Fall 2025, the Spark! team focused on improving grading consistency for text-based short answer assessments using Azure AI Foundry. Through prompt iteration and testing with Mike’s teaching team, the AI-generated grades and feedback became acceptable by the end of the semester, and the current prompt/model configuration does not require major changes for text-only assessment grading at this time. However, the team also identified a key limitation in Azure AI Foundry’s multimodal capabilities. While Azure can extract text from PDFs and slides, including text embedded in images when preserved in the file, it cannot reliably interpret or summarize images that do not contain text. Since CS 581 includes course materials and assignment submissions that are often Excel-based and diagram/image-based, this limitation affects the feasibility of expanding the AI grading workflow beyond text-only assessments. This semester, the project will continue the grading pilot while focusing on selecting and implementing an AI platform that can reliably handle multimodal course materials and assignment submissions (PDFs with images, Excel files, and image/diagram submissions), with API access as a core requirement. Client Note: We want to parse Blackboard data (student submissions, lecture material, previous semester student submissions, etc.) in a way that a RAG system can read. Our main concern is that existing files are not readable using RAG, either in Azure or other tools. Whatever tool they choose (if Azure is not workable for our assignment use case), if needed please create a script to convert all lecture and student submissions to outputs interpretable by rag. This may need to be word documents or text. This project can be divided into 3 phases: Phase 1: Research and Tool Comparison (Multimodal Benchmarking) The team will evaluate multiple AI platforms/tools (Azure and non-Azure options) to determine which best supports CS 581 needs, especially multimodal interpretation. To keep testing consistent, ETI will provide three representative CS 581 files: • One Excel file (example student assignment submission) • One image submission (diagram, exported as PNG/JPEG) • One lecture PDF (“Principal Lectures”) containing a mix of text and images The team will run a structured comparison across tools, documenting: • whether the tool can read and interpret each file type correctly • whether it can use course materials as reference during grading • API access and integration feasibility (and rough cost considerations) Phase 2: Environment Setup and Configuration After selecting a platform, the team will: • set up a local testing environment using that platform’s API • implement a repeatable workflow for ingesting and evaluating the 3 sample files • document setup steps so the system is reproducible If Azure remains the chosen platform, ETI will provide documentation for the existing system so the team can mirror the current configuration as closely as possible (then iterate based on research findings). Phase 3: Pilot Support and Iteration (Live Course Data) Once the environment is running, the team will support active CS 581 pilots by: • continuing to improve grading for short-answer assessments • expanding analysis into discussion posts (if in scope for the pilot) • configuring and testing grading support for Excel assignments and diagram/image submissions • comparing AI outputs to facilitator/instructor grading and documenting improvements over time • recording any changes made to the environment, prompts, rubrics, or workflow Stretch / add-on within Phase 3: Develop clearer evaluation metrics for grading quality/consistency (potentially informed by IRR-style approaches and ETI research), and propose how to adopt them.

BU MET College

Biking Project

BCU Labs is building data-driven tools to support evidence-based transportation advocacy and safer cycling infrastructure in Greater Boston. A large volume of open transportation data already exists — including bike-share ridership records, municipal bike counts, and crash databases — but these datasets remain fragmented, inconsistently formatted, and difficult to use for longitudinal analysis or public-facing tools. This project aims to bridge that gap by creating a unified, analysis-ready transportation data infrastructure that consolidates these disparate sources into a centralized master database. Student teams will develop automated pipelines to ingest and standardize monthly Bluebikes ridership data, reconcile unstable station identifiers and location metadata, and integrate external bike count and crash datasets from city and state agencies. With a stable and continuously updated data foundation, students will enable new interactive visualization and analysis layers on BCU’s existing Bicycle Stress Map — including ridership trends, safety-risk overlays, and infrastructure gap indicators. In doing so, the project will support data-driven trend analysis, safety assessment, and transportation planning insights, while making complex transportation data more transparent, accessible, and actionable for community members, planners, and policymakers.

BCU Labs

Boston Heating Systems

To help reduce building-related greenhouse gas emissions, BCAN conducts targeted outreach to residents and property owners who have the greatest potential to reduce energy use. A critical first step in this effort is identifying what type of heating fuel or energy source each building uses, since heating systems are a major driver of emissions. Currently, there is no centralized public dataset that clearly identifies the heating fuel used by individual buildings in Boston. However, several publicly available datasets contain partial or complementary information that, when combined, can be used to infer or predict a building’s heating system. These include the Boston Tax Assessment Database, which contains building-level heating type information, siding, roofing, and construction dates, and the City of Boston Approved Building Permits, which include records of heating system installations and upgrades. This project aims to integrate and reconcile these datasets to infer the most likely heating system used by each building. For buildings with missing or incomplete information, we will use overlapping attributes across datasets to estimate the heating system type using the predictive logic provided below.

BCAN

Boston Voter App

Yawu Miller is founding a new nonprofit news agency to cover the BIPOC communities in Massachusetts. As part of this effort, he is developing the Boston Voter Web App to address the lack of accessible information around voting logistics and candidates in municipal elections in Boston. The project aims to create a web application where Boston voters have all vital voting information in one central location in service of the ultimate goal of increasing BIPOC voter turnout in municipal elections. Many news organizations provide articles or voting guides around the election, but they are often difficult to find, last minute, and represent a narrow perspective around candidates. Additionally, most election resources are around larger races with little coverage or few resources available for local elections, where candidates have the potential to impact people’s lives more directly. BIPOC voters often have the lowest voting rates especially in municipal elections for a variety of reasons including lack of information and logistical barriers such as hours of work, and family duties, that prevent them from reaching the polls. This project aims to remedy that for BIPOC voters in Boston. The principal goal of this project is to connect the following resources into one place: 1. Who is my Boston City Councilor? 2. When, Where, and How Do I Vote?

Yawu Miller

Bus Delay Patterns in Boston

Boston’s buses experience frequent delays caused by double parking and curb-use conflicts. While these delays are widely recognized, existing data does not clearly show where slowdowns recur or how they accumulate across time and space. This project centers on visualizing recurring bus slowdowns using GTFS-realtime data and related indicators such as 311 complaints. Students will design map-based and temporal visualizations that make slowdown patterns legible at the route and corridor level. The focus is on revealing spatial concentration, time-of-day patterns, and alignment with curb-use signals—not on optimizing detection models. The visualizations are also intended to highlight recurring patterns and correlations. Final outputs will help LivableStreets communicate where bus delays consistently occur and support advocacy for targeted curb management or enforcement.

Livable Streets

Campus Planning

BU currently distributes emergency guidance through emails, websites, and static plans—resources that can be difficult to locate or act on during high-stress moments. This project will build a scenario-based mobile app that delivers step-by-step emergency instructions (active threat, fire, severe weather, evacuation) in a format that's easy to access and follow in a crisis. The app will pull from existing BU emergency plans and protocols rather than creating new content, repackaging trusted information for mobile delivery.

CISS Native Ads

The Fossil Fuel Native Advertising Observatory project supports the development of an automated, research-driven system for analyzing how fossil fuel companies use native advertising to shape public narratives around climate and energy. During Fall 2025, the team focused on evaluating and extending a claims-based classification model for identifying rhetorical strategies and misinformation-related themes in fossil fuel–sponsored native advertisements. The system was tested on a growing corpus of sponsored articles collected from major news outlets and labeled using a custom claim taxonomy. By the end of the semester, the team validated the claims-based approach against both sentence-level and article-level annotations, demonstrating that this method captured nuanced messaging patterns that prior CARDS-based approaches failed to identify, particularly around false solutions such as carbon capture and petrochemical framing. For more notes on key findings from the previous semester please refer to the “Key Findings from Fall 2025” section. This semester the team will contribute to the development of CLAIMS 2.0, an expansion of the original CLAIMS model that is more claims based and less reliant on keywords. The goal is to use an ‘open-coding’ approach to typology development using a set of ‘seed claims’ about green technologies (such as CCS and renewable energy) and fossil fuels (such as oil and gas), to arrive at an exhaustive set of claims made about these fuels/technologies in native ads. The results will contribute to the larger CLAIMS 2.0 model which detects claims about a range of green technologies and fossil fuel products. Future teams should validate outputs against hand-labeled examples and clearly communicate model limitations alongside results.

Michelle Amazeen

CivicChat

CivicChat helps citizens learn about local to national government and make more informed civic decisions, in their native language, simply by talking to their phone. We want to expand this project’s vision, to help incentivize and engage voters in local Boston elections. Along with an accessible chatbot, we want this platform to utilize public data to help tailor information to the user, and guide them through the voting process. Another older Spark! Initiative aimed to do this very same thing, https://bostonvoter.com/, and we want to pull features from here to this new civic chat platform. This will include the feature under the “your voter info” tab, drop box locations, and possibly the ballot form.

Blackfacts

Constituent App

The goal of the District 4 app is to provide a platform for citizens of District 4 to access relevant information about their district including: resources, announcements, opportunities for civic engagement, and sharing events. This semester will focus on translating design concepts into a functional and user-friendly application that supports the needs of both District 4 constituents and the city council office. • Fleshing out admin side • Implement figma designs from the Fall and expand desktop view • Develop and Integrate content + designs for the calendar feature • Develop and Integrate content + designs t for resource library feature • Integrate District 4 social media with the web app for easy diffusion of district updates Phase 1 (Fall ‘24): • Build a strong foundation for the app by developing the calendar feature and researching strategies for effective content integration Phase 2 (Spring ‘25): • Building the calendar functionality to ensure it is fully interactive and easily maintained • Develop the resource library to display resources for District 4 citizens • Implement social media integration to automate the sharing of updates • Testing and refining app features to improve usability and align with user needs Students will focus on improving the app by iterating on the implemented designs and features, incorporating feedback to refine the calendar and social media integration for better functionality and user experience. Overall, the goal is for constituents to easily access information about their district and for the city council office to efficiently update and populate the site with content.

D4 City Council

Data Common

This project is a continuation of work initiated in the previous semester to analyze and improve data visualizations for Housing Navigator Massachusetts’s website, with a focus on understanding how users engage with the affordable housing search engine and property listings on housingma.org. Building on an existing analytics and dashboarding foundation, this semester’s work will deepen, refine, and extend prior analyses rather than starting from scratch. In the previous phase, student teams used Google Analytics (GA4) and Looker Studio to explore high-level user engagement patterns across the platform. Initial findings surfaced meaningful differences in usage by weekday vs. weekend, hour of day, and device type, and helped identify opportunities to improve clarity and usefulness of existing dashboards. During the Spring 2026 semester, the project will focus on enhancing and operationalizing these insights to better support Housing Navigator’s internal decision-making and external reporting needs. Key areas of emphasis include refining geography-based analyses (e.g., districts, regions, or planning areas within Massachusetts), improving dashboard filters and usability, and extending analysis of internal publishing activity now that readable organization- and role-level identifiers are available.

Abby Werner

Deportation Data

This Data Days 4 Good project invites students to build an interactive, searchable dashboard that helps GBH journalists and the public better understand foreign-born population trends in Massachusetts Gateway Cities. The goal is to enable users to explore how immigrant populations have changed since 2010, understand demographic shifts at granular levels, and identify patterns related to economic assimilation. Students will create a data-driven tool using Census data that allows users to look into any Gateway City and see how the foreign-born population has changed over time, broken down by specific countries of origin (e.g., Vietnamese, Brazilian, Lebanese, Chinese) rather than just top-level categories. All data will be presented as rates and per capita to enable apples-to-apples comparisons across cities and to statewide totals. The dashboard will include interactive filtering and comparison features, allowing journalists to identify which cities have the highest or lowest rates of specific variables, compare cities side-by-side, and generate city profiles. A key feature is a chatbot interface that enables users to ask questions of the data and generate graphs dynamically, making the data exploration accessible to non-technical audiences. This tool will support GBH's investigative reporting by surfacing patterns about which cities have experienced the greatest foreign-born growth, which populations are growing fastest, and indicators of economic assimilation including median income, housing patterns, educational advancement, and employment.

GBH

Domestic Violence in New Hampshire

Women in New Hampshire are experiencing domestic violence at significantly varying rates compared to the rest of the United States — they were once above the United States’ count, but are now drastically below it. This project seeks to evaluate domestic violence incidents over time in New Hampshire and determine how many, where, what types, when, and what the impacts were. We have data on domestic violence incidents going back to 2000. You will also explore and evaluate other government datasets to see if/how they could be valuable for the project.

Melanie Plenda

Domestic Violence in Police

Boston 25

BU Campus Planning and Operations

Emergency App

Center for Career Development

Employer Engagement Analysis Project

The Employer Engagement Analysis Project aims to help the Boston University Center for Career Development (CCD) better understand how employers interact with BU students and offices through the university’s career ecosystem. Using data collected from Handshake between Spring 2020 and Summer 2025, this project will analyze five years of employer activity, including job postings, event participation, and student applications, to identify patterns in engagement and hiring behavior. By integrating datasets that include employer information, event records, job posting data, and student participation, the team will explore which employers are most active, which engagement activities generate the highest student interest, and which factors are most predictive of successful hiring outcomes.

Data Science / Data Visualization · Spring 2026

Federal Grant Cancellation Tracking Project

The current administration has, in many cases, quietly reduced or terminated funding, often without public announcements. This lack of transparency makes it challenging to track the full scope and impact of these cuts. This project aims to systematically identify federal grants and awards to Massachusetts that have been cancelled or rescinded since the start of the current administration in January 2025. By establishing an automated detection process, the Senator’s office can quantify the total financial impact, determine which agencies and programs are affected, and respond quickly to protect critical funding and work with affected organizations and communities in Massachusetts. Funding modifications will be classified into two categories: • cancelled grant: An award made by the federal government where funds were never transmitted to the awardee, or the grant was halted before the full amount was disbursed (since federal grants are often paid in installments). • Rescinded grant: An award where funds were transmitted to the awardee but later clawed back by the federal government. The project will use the USAspending API to analyze each grant’s award history in the dataset. Research indicates that certain award modifications, such as setting the amount to $0 with an action type of “revision”, can indicate a cancellation. Rescinded grants can be identified using criteria where funds have already been disbursed but are later withdrawn, which is typically reflected as a negative transaction amount.

Liam Horsman

Fiscal Health Analysis System

DevTech Systems seeks to build an automated fiscal health analysis system that eliminates the current manual process of extracting financial data from local government reports. These reports (CAFRs/ACFRs) vary widely in format and structure and may appear as PDFs, Markdown files, or JSON exports. The goal is to create a scalable pipeline that can reliably identify, extract, normalize, and analyze key financial components—such as assets, liabilities, revenues, expenses, fund balances, and debt measures—across multiple localities and years.

DevTech Systems

Flooding Insurance

This project will investigate how climate change–driven natural disaster risk is reshaping the home insurance market in Massachusetts, leading to rising premiums, coverage reductions, and policy non-renewals. The team will analyze state insurance data, census demographics, and historical disaster records to identify geographic and socioeconomic patterns in insurance rate changes. A central focus will be examining how these shifts disproportionately impact lower-income and environmentally vulnerable communities, potentially limiting residents’ ability to maintain homeownership and financial stability. The findings will support reporting on environmental justice implications, industry response to climate risk, and the growing financial burden placed on the state as the insurer of last resort.

Commonwealth Beacon

Food Access

The Food Access team at La Colaborativa has collected a substantial amount of data on community members who access its food programs, including pantry visits, home deliveries, El Mercado participation, and mobile produce truck usage. This project aims to transform that sensitive but powerful data into clear, ethical, and compelling data visualizations that communicate who is being served, where services are reaching, and the overall impact of the Food Access program. Through this project, the team will support La Colaborativa in developing data visualizations that can be used for both external audiences (funders, conferences, partners) and internal decision-making. The work will emphasize storytelling, equity, and responsible data use, helping the organization better articulate community needs and program impact while protecting participant privacy.

La Colaborativa

Fund Tracking

The current administration has, in many cases, quietly reduced or terminated funding, often without public announcements. This lack of transparency makes it challenging to track the full scope and impact of these cuts. In the initial phase of this project, the team built a systematic approach to identifying federal grants and awards to Massachusetts that have been cancelled, rescinded, or otherwise modified since the start of the current administration in January 2025. Using a combination of USAspending data, background research, and exploratory analysis, the team developed and validated a custom classification framework capable of distinguishing between cancelled grants (where funds were never fully disbursed), rescinded grants (where funds were disbursed and later clawed back), partial rescissions, and administrative adjustments. To operationalize this work, the team built an automated ETL pipeline and delivered a Streamlit dashboard that provides a clear, searchable view of when, where, and how federal grant cuts occurred across Massachusetts. This semester, the focus will shift from initial detection to deepening coverage, improving comparability, and strengthening temporal analysis. The team will extend the existing dashboard to incorporate richer award-level data from the USAspending API, including visual indicators that flag newly discovered or previously unknown awards as they appear or change over time. Additional visualization layers will be developed to enable state-by-state and tract-level comparisons, allowing users to examine how federal grant cuts vary across political, social, and economic contexts. Finally, the project will introduce time-series visualizations that track the full lifecycle of awards, highlighting delays, revisions, or gaps in USAspending data updates and making the timing of funding changes more transparent.

Senator Ed Markey

Progressive Massachusetts

Funding Patterns Project

This project examines how campaign donations flow to Massachusetts state legislators and how funding patterns relate to legislative influence, committee power, ideology, and election outcomes. Using campaign finance data from the Massachusetts Office of Campaign and Political Finance (OCPF), the project will link donors to individual legislators and analyze contribution patterns across committee memberships, committee chairmanships, and legislative leadership roles. The analysis will investigate whether legislators holding influential positions receive greater financial support, which industries and donor groups target specific committees, and whether donations are primarily local or national in origin. In addition, the project will merge campaign finance data with legislator party affiliation, ideology indicators, district-level election results, and donor geographic data to explore whether donor behavior can help explain or predict electoral success. Where data allows, the project will further analyze donation patterns in relation to key policy areas, including prison and criminal justice reform, health policy, education, housing, public transportation, and energy and environmental policy. The goal is to provide a clearer, data-driven understanding of how political money is distributed in Massachusetts and how it aligns with legislative power structures and policy priorities.

Geotag Tree (One Acre Fund)

This project aims to support One Acre Fund in scaling its tree census and monitoring efforts through a low resource computer vision assisted system for identifying and tracking young trees in dense, non grid planted agroforestry plots. Enumerators currently collect geotagged images, record species, and manually enter planting year, but GPS imprecision, closely spaced trees, blurry images of seedlings, and lack of consistent scale make reliable identification and diameter estimation difficult. The project explores whether computer vision models can reach confidence levels high enough to be useful at scale, potentially leveraging visual cues such as painted markings, QR tags, or low cost physical identifiers introduced after a survival threshold, given that tagging all seedlings upfront is impractical. The work will evaluate tradeoffs between accuracy, cost, and time efficiency across sensing approaches such as standard smartphone photography, QR based scale references, and emerging mobile LiDAR on select consumer devices. With hundreds of trees per acre, multiple species per country, and survival rates of 40 to 60 percent, the project focuses on identifying field ready tools and workflows that meaningfully improve data quality while remaining feasible in low resource, high volume environments.

One Acre Fund

Granite State News: NH Profiling

Granite State News

Haynes Construction AI Agent

Haynes Construction is a commercial construction company with over 20 years of experience delivering large-scale projects across retail, hospitality, and multifamily sectors. The company manages multiple datasets spread across four main shared drives, along with SaaS platforms including Procore (project management, open API), FollowUp CRM (sales pipeline management, open API), and Sage100 (financial system on SQL, no cloud-based API). Despite the wealth of historical data, information is siloed and inconsistently indexed, creating inefficiencies in data retrieval and analysis. The goal is to unlock the potential of this data through AI-driven analytics and a chatbot interface to improve decision-making, forecasting, and operational efficiency. This project focuses on developing an AI Agent to make data retrieval and analysis more efficient and intelligent. The Haynes AI Chatbot will serve as a data assistant for construction project insights, allowing users to ask natural-language questions (e.g., “What was the cost per trade for Project X?” or “Compare scheduled vs. actual completion dates for 2024 projects”) and receive instant, data-driven answers. This first phase focuses on uncovering and structuring five foundational datasets that the chatbot can query, enabling predictive and operational analysis in future phases.

Haynes Group

· Spring 2026

Herbaria

The Harvard Herbarium has an archive of thousands of plant specimens. These specimens are made up of pressed and dried flowers, leaves etc on a large paper, along with an annotated note about their species, collection date etc. These specimens are digitally scanned and stored in a system that allows researchers to view scans and manually enter the information to digitize the records. In addition, the records are searchable and viewable to other researchers.The herbarium has designed several custom solutions that speed up this manual data entry process. For this project, they would like to adapt these workflows to work with an industry standard system called Symbiota. This is used by other Herbariums around the world, each with their own custom instance.

Harvard Herbarium

Homeward

ULI wants to digitize Homeward's scoring system, currently a multi-tab Excel spreadsheet that interrupts gameplay and confuses participants. This UX project focuses on researching how players and facilitators interact with the current tool, identifying pain points, and designing a more intuitive digital interface. The team will conduct user research, map the existing workflow, and create a validated prototype for a web or tablet-based scoring tool. This work will set the foundation for a future SWE team to build the final product.

ULI

BU Community Service Center (CSC)

Hours Tracking App

For multiple semesters, the CSC has been working with Spark! to improve their online. Platform. The platform has 3 main users: (1) Student volunteers, (2) BU CSC Admin, (3) nonprofits/3rd-party organizations. The admins of this platform are responsible for large amounts of tracking and management that can be streamlined with more efficient user flows and features.

Human Trafficking

Attorneys, advocates, and policy advisors working on human trafficking issues currently rely on fragmented sources—peer networks, agency websites, broadly shared emails—to stay informed about legal and policy changes. This project aims to design a centralized, searchable resource that aggregates updates across legislation, regulations, case law, and enforcement actions. The UX team will design an interface that makes this information easy to find and navigate. Users include attorneys, policy advocates, survivor support organizations, and law students. Key design challenges include balancing comprehensive filtering with minimal click depth, supporting multiple user types with different needs, and incorporating trauma-informed features like quick exit buttons.

BU Law

Indigenous Architectural Reconstructions in Palestine

Many historical buildings have been lost or altered over time, leaving only fragmented visual records in photographs or paintings. To digitally preserve architectural heritage, we need methods that can reconstruct buildings from limited visual inputs. While modern computer vision has made progress in generating multiple views and 3D reconstructions from single images, applying these techniques to historical imagery presents unique challenges such as image quality, perspective limitations, and missing data. Last semester, a Spark team built an end-to-end proof-of-concept single-image 3D reconstruction pipeline. The system generates multiple novel views of a building from one photo, estimates depth for each view, converts depth maps into point clouds, and merges them into an exportable 3D model. This semester, the team will build on that foundation to improve quality, reliability, and usability of the reconstruction outputs. The focus for this semester will be on developing and testing the image generation models, while the long-term goal is to use these generated views as the basis for forming 3D building models. If more images are needed, we also have a ready-to-use Google API script that can collect building images from multiple angles based on geographic coordinates. What is the possibility of expanding the scope to the neighborhood, multiple buildings, street view, and diverse urban forms? While capturing its evolution through time? Based on images such as maps and aerial images, By the end of the semester, the team will aim to: • Improve control and consistency of AI-generated novel views. • Reduce geometry artifacts and improve mesh / point cloud cleanliness. • Increase robustness to real-world images (clutter, multiple buildings, imperfect framing). • Move toward a more automated, seamless workflow from input image to exportable 3D asset.

Osama Alshaykh

Kingdom App

Kingdom is a gamified platform that ranks independent artists by location and category using engagement data from external platforms (like Spotify and Youtube) and in-app activity. The goal of this project is to design a clear and engaging user experience foundation for the Kingdom app, focusing on discoverability, ranking transparency, and artist exposure at the local and regional level. The goal of this Spark! The project is to design the UX for the Kingdom App, focusing on artist onboarding, profiles, rankings, discovery, and gamified recognition. The student team will translate the client’s vision into a clear, scalable user experience that communicates how rankings work, how artists compete, and how fans discover emerging talent. The project will culminate in a high-fidelity, clickable prototype and supporting documentation that can serve as the foundation for future development beyond Spark!.

Onyx White

Legal Agent/Chatbot

This project is a continuation of work from last semester, where students built a proof-of-concept AI chatbot to answer housing-related legal questions using a Retrieval-Augmented Generation (RAG) pipeline. The existing system ingests curated legal sources and returns natural-language answers with citations. In Spring 2026, the student team will focus on improving, evaluating, and validating the existing chatbot rather than building a new system from scratch. The work will center on three tightly scoped phases: 1. Model & Data Improvements • Expand and refine the chatbot’s knowledge base using actual legal text (e.g., statutes, regulations, and selected primary sources), not just metadata or secondary summaries. • Improve document embeddings, retrieval quality, and citation grounding. • Incorporate additional vetted legal Q&A pairs developed by clinic students. 2. Evaluation & Testing Framework • Build a gold-standard test set consisting of legal questions, expected answers, and required citations. • Implement structured evaluation using frameworks such as RAGAS, Phoenix, or similar tools to measure retrieval accuracy, citation quality, and response consistency. • Design a system to log and track model outputs over time, enabling week-over-week comparison and iterative improvement. 3. Benchmarking & Comparison • Where permitted, compare the Spark-built chatbot’s responses against outputs from Harvey AI, a legal AI platform available to BU Law. • Analyze differences in accuracy, sourcing, explainability, and cost tradeoffs. • Use insights from benchmarking to inform recommendations for improving the open, lower-cost system. The goal is not to replace legal advice, but to create a transparent, well-evaluated tool that helps tenants understand their rights and prepares them to engage more effectively with legal systems.

BU School of Law

Innovation · Spring 2026

LUCE

This project will expand beyond the initial release by incorporating advanced functionality designed to improve data analysis, accessibility, and advocacy impact. Future development will focus on creating a dashboard and statistics view that highlights enforcement patterns and frequency across Massachusetts. Tools for manual-to-digital conversion of scanned documents and automated PDF parsing will streamline how information enters the database, reducing reliance on manual entry. Partners such as the ACLU and NIAJ will be able to request scenario analyses, enabling more strategic advocacy efforts. Additionally, the platform will integrate improved analytics and visualizations, helping community members, legal teams, and policymakers better understand trends and respond effectively. This initiative is inspired by ICEWatch, a similar tool created by the Immigrant Defense Project used to track ICE enforcement activities in New York.

Sarah Sherman-Stokes

MassCourts V3

This project aims to develop an application to provide searchable access to millions of court cases across Massachusetts. This project is building the next generation of a previously existing web application built by the client, Civera, which is currently hosted in WordPress. The client, Civera, manages an extensive dataset of Massachusetts court cases collected through their scraping tool. The goal is to transition from raw scraped docket results to a structured, normalized database integrated with a functional front-end search engine.

Civera

Matching Higher Education Institutions

This project builds a system to standardize university names across multiple higher education datasets by matching them to unique institutional identifiers (UNITIDs). Students will develop and refine a fuzzy matching algorithm that automates this process, using a manually curated training dataset of institutions. This standardized data environment allows Boston University (BU) to be accurately compared with peer institutions across national higher education datasets. Students will investigate what the standardized datasets reveal about where BU is most impactful compared to other universities – particularly in research activity and grant funding. With additional data sources (such as research funding databases and OpenAlex affiliation records), students will explore BU’s strengths across academic areas and departments.

Philip Lindsay

Microcredentialing Project

The goal is to create a digital platform (web-based) that helps students learn and get evaluated on core chemistry lab skills through videos, quizzes, and in-person or video-recorded demonstrations including a dashboard for instructors to track progress and give feedback. This project is designed to transform how first-year chemistry labs are taught, with the long-term goal of changing how skills-based micro-credentialing is done. Students will use an app that helps them learn important lab skills through question-embedded videos, formative quizzes, and then get guidance about hands-on practice and assessment rubrics. After practicing in the lab, students will go to a teaching assistant (TA) who will check if the student has mastered the skill. Using the provided rubric, the TA will give them targeted feedback on each element of the skill and then overall rating of “proficient” or “still learning.” The system will track progress of each skill and provide a micro-credential report for students and instructors. This system will make learning new skills more interactive as well as providing students with more helpful and personalized feedback, rather than just a simple pass or fail. Overarching goals • Enable students to learn lab skills asynchronously through videos and quizzes • Create an AI-assisted robust evaluation process Track student progress and skill + badge status in a centralized system • Export credentials to external platforms (e.g., LinkedIn) Current Problem Proposed Solution 800+ students tracked in Google Sheets Build centralized system with student profiles + progress No written feedback or rubric given in advance Embed rubrics into app and share ahead of lab Zero/One grade in Blackboard Replace with granular badge status and comments TAs multitasking safety + grading Consider camera setup or partner-recorded submissions No record of attempts Show history of attempts and feedback for each skill No feedback loop into curriculum Allow instructors to see macro-level pain points

Binyomin Abrams

Misdemeanor Measurement

Misdemeanor arrests and convictions are the most common form of contact with the U.S. criminal legal system, yet their long-term impacts are routinely overlooked in research and policy debates that focus primarily on felony convictions. Although legally classified as low-level offenses, misdemeanor convictions can create lasting criminal records that restrict access to employment, housing, education, and professional licensing. Because misdemeanors account for more than three-quarters of all criminal cases filed annually, failing to measure them obscures the true scale of criminalization and underestimates the population that stands to benefit from record-clearing reforms. The Clean Slate Initiative (CSI) has developed a unique data model to estimate the prevalence of criminal records, including an innovative methodology for quantifying misdemeanor convictions using FBI crime and arrest data. However, the FBI’s 2021 transition from the Uniform Crime Reporting (UCR) System to the National Incident-Based Reporting System (NIBRS) requires a rigorous methodological update. This project will integrate NIBRS into CSI’s model while accounting for under-reporting and missing data, ensuring accurate estimates of misdemeanor records. Strengthening misdemeanor measurement is critical to CSI’s mission, as Clean Slate policies most often seal misdemeanor convictions and rely on precise estimates to demonstrate their reach, equity impacts, and life-changing potential. Students that partake in this project will gain experience learning about the criminal justice system, the national standard for crime reporting, manipulating and cleaning relational datasets, and modeling techniques to account for missing data.

Clean State Initiative

Mobile App Dev

Barterloo is rooted in the ancient, time-tested practice of barter, which supported communities for thousands of years before money existed. What we're building isn’t something new. It’s a remembered system of value exchange that honors what we each have to give, outside of traditional currency. With today’s technology, we have the tools to reawaken that system on a global scale. We believe no one should be living in lack when we all hold value in the form of time, skills, services, and secondhand goods. Barterloo makes it easy to exchange those offerings through a community-centered platform designed for both individuals and small businesses. Our goal is to build a self-sustaining, community-governed ecosystem where people can connect not just for what they want, but for what they truly need. It’s time to move beyond competition for limited resources and start building longer tables that create equity, access, and mutual support across the world.

Kelly Dempsey

SWE / UX · Spring 2026

MOLA

The goal of this project is to build a comprehensive resources guide that will provide individuals in the LGBTQIA2S+ community with important resources, services, events, and other forms of support in the City of Boston. For Fall 2024, this project will begin with a draft of a multi-layered map that encompasses the location of organizations in respective neighborhoods.

Kimberly Rhoten

Earl Center for Innovation

Music App

This project is a pilot, community-engaged collaboration in which BU undergraduates enrolled in DS488 partner directly with students from O’Bryant School of Mathematics & Science, under the guidance of Michael Chang, to co-design a map-based, playlist-driven digital experience based on movement through the city. This project is unique in the sense that it is UX research intensive with low visual design opportunities. Rather than positioning high school students as end users, this project is grounded in deep collaboration and co-creation. High school students act as design partners throughout the process, contributing to research, ideation, and evaluation. In addition to partnering with high school students, the project will include collaboration with a BU Wheelock College of Education student who will bring expertise in learning sciences and youth-centered pedagogy. Student teams will meet weekly in person at O’Bryant High School every Wednesday at 9:00am, with additional intensive work sessions during the semester. Participation requires a high level of engagement and commitment from all students involved. The initial design challenge will focus on a map-based digital experience guided by the prompt: “Create something for someone you care about.” One proposed concept explores a tool that helps a younger sibling or peer navigate Boston while allowing community members to curate location-based playlists, stories, or reflections—creating an accessible entry point for examining how data, algorithms, and AI shape everyday experiences. During Spring 2026, the scope of DS 488 will prioritize research, strategy, and low-fidelity UX design. The primary goal is to prototype the UI/UX research/experience rather than produce a fully developed product. This pilot will lay the groundwork for continued development in subsequent semesters and ultimately evolve into a new course. Student Roles Wheelock Student: • Mindful of Pedagogy, thinking about how the team can teach the O’Brytant students together Spark! College students • Demonstrating learning outcomes through teaching and mentoring, as well as doing. Highschoolers • Learning from college students, receiving direction and mentorship • Contribute to deliverables and user research Meeting schedule • Weekly 8.30am -10.00am meetings on Wednesdays (50min), every week. • Evan pays for uber with spark! card (Nolan slack lydia), • Funding for transportation (Nolan to reach out to ziba about this) some budget for this but coming out of entire class expenses, so we are willing to split it. We tell you what it cost and he sends the other half after the semester • Evan to own coordinating this : They all meet cds or agreed location and one uber takes them to the highschool • People to include: evan, undergrad team, highschool, Jorge (lead), Micheal, Omar (optional) • Evan to fill out Query form by tomorrow • 55 Malcolm X Blvd, Roxbury, MA 02120 • Teaching and doing with the high school students • Weekend team meetings/intensives (4-5 hr) • Evan or Micheal to coordinate this meeting • Extension of work sessions or work that is more focused on some of the ethical issues emerging • 55 Pilgrim Rd, Boston, MA 02215 • People to include: evan, undergrad team, highschool, Jorge , Micheal (lead), Omar (optional) • 3 in entire semester (there is a syllabus for this) • Undergrad only meetings once every other week with PM (evan) and Jorge • Evan to own coordination • People to include: Jorge, evan, undergrad • Optional invite to Morgan and Mihyun • Zoom and/or in person There will be several weekend intensives adding up to 18 total hours throughout the semester. Students (including evan) will be compensated $15 ($270 total) an hour for these intensives.

ODLS Tracker

This project supports immigration legal clinics with two connected monitoring tools, built in sequence. The first tool (ODLS Tracker) helps lawyers locate detained clients and preserve evidence when ICE fails to update public records. Students will build an automated system that checks ICE's Online Detainee Locator System for specific A-Numbers, captures timestamped screenshots, logs results in a database, and sends alerts to attorneys when a detainee appears or their location changes. The output is a web dashboard where legal teams can track detainees in real time and maintain searchable evidence logs for habeas petitions and deportation defense. The second tool (Visa Bulletin Monitor), to be completed if time permits, addresses long-term case tracking for SIJS youth and other visa applicants facing multi-year backlogs. Students will build a system that parses the monthly State Department Visa Bulletin, compares cutoff dates against client priority dates, and notifies lawyers and clients when cases advance, become current, or retrogress. The output is a "tickler" system that reduces manual tracking and prevents missed filing windows. Both tools share similar architecture—automated data ingestion, alert pipelines, and simple attorney-facing interfaces—making them natural complements within a single semester.

Cornell Law School

OneScor Connector

OneScor (What you’re building for) OneScor is an enterprise risk and resilience platform built by Iroko Technologies. It helps organizations understand how their many interconnected systems, applications, and vendors depend on one another so they can better assess risk and make informed operational decisions. In modern enterprises, data lives across hundreds of systems with different schemas, naming conventions, and formats, making data alignment and integration one of the most expensive and error-prone parts of adopting new platforms. Integration Platform (What you’re building) This project prototypes an AI-assisted schema-matching integration workflow for OneScor. Students build a web-based experience that connects to third-party systems, retrieves table schemas and sample records, and uses ReMatch-style AI matching to suggest how external fields map to OneScor’s data model. The workflow allows users to review, adjust, and validate mappings before simulating secure ingestion into OneScor’s staging environment, demonstrating a clear and user-friendly approach to enterprise data onboarding.

Iroko Technologies

BU Center on Media Innovation for Social Impact

OnThePorch

The Public Safety and Community Sentiment project aims to help Dorchester residents and community leaders access timely, trustworthy information about local public safety issues and community events by combining official city data with verified community reporting. During Fall 2025, the student team delivered a working AI-powered community assistant that integrates a chat-based interface with a filterable event calendar. Users can ask natural language questions such as “What’s happening this week?” or browse upcoming events directly from a sidebar. The system currently ingests data from Dorchester Reporter, Boston.gov APIs, community newsletters, and uploaded documents, with automated daily updates. A key focus of Fall 2025 was improving data ingestion, source coverage, and usability rather than expanding analytical complexity. The team implemented automated pipelines for newsletters and public data, ensured clear source attribution with links to original reporting, and designed a UI that was approved by the client for clarity and accessibility. Client feedback emphasized the importance of avoiding AI editorializing, particularly when discussing sensitive public safety incidents. The preferred interaction model is to present verified facts first, then optionally offer historical trends or sentiment context only when explicitly requested. This insight reframed the system’s role as an assistive information tool, not an interpretive or opinionated one. While the current system functions as a proof of concept, several known limitations remain, including incomplete shooting incident location data, gaps in recent datasets, and the need for structured user testing with community members. This semester’s continuation project will focus on strengthening reliability and trustworthiness, expanding newsletter ingestion, and developing a clearer framework for separating factual reporting from trends and optional context. The intent is to conclude the student-driven development cycle with a more robust prototype that can be evaluated through user testing and positioned for potential productization or handoff to professional development resources. In addition to newsletter ingestion, this semester will extend the platform to support community conversation audio ingestion. The team will prototype a lightweight backend/admin upload interface where a trusted partner can submit an audio file and required metadata (meeting date/time, location, context, agenda, and any official roles present). The system will then generate a transcript, redact personally identifiable information for community participants, and index the content so that the assistant can answer questions by quoting the transcript explicitly (for example, “A community member said…” with a timestamp/date) rather than mimicking a community voice. This work supports the project’s core trust requirement: the assistant should provide verifiable information with clear sourcing, and only use transcript-based sentiment or context when the user explicitly requests it. (Please confirm with client before continuing this part) Project recommendations & next steps: (Ideal Output & Final Deliverables) Recommended Next Steps • Reintegrate and validate shooting and shot-spotter data with accurate latitude/longitude information. • Implement a prompting framework that clearly separates: • factual reporting • historical trends • optional sentiment or contextual analysis • Expand automated ingestion to additional community newsletters identified by the client. • Conduct usability testing with at least 20 community members to assess trust, clarity, and usefulness. • Add power-user functionality such as CSV exports and structured summaries for reports. • Building a mechanism to ingest/upload new community conversation audio files, similar to how newsletters are ingested. • Capturing required metadata at upload time (date/time, place, meeting context, agenda if available, whether public officials were present, etc.). Things to Avoid • Do not allow the AI to automatically editorialize or blend historical quotes with current incidents. • Do not surface internal spreadsheets, raw transcripts, or documents containing names. • Avoid overloading responses with sentiment when users are asking for simple factual information. Common Blockers • Incomplete or delayed public datasets (especially for recent years). • Sensitivity around how incidents and quotes are framed. • Token and context limitations when working with large text corpora.

Phoenix Engage Clinical

Phoenix Engage Clinical is a workflow automation platform that helps sponsors and clinical sites move faster through site identification, feasibility, document collection, contract negotiation, budget negotiation, and site startup for clinical trials. Phoenix focuses on the operational and administrative work that happens before a site can enroll patients, with the goal of reducing delays, manual effort, and back‑and‑forth communication. Phoenix uses automation and AI assistance to: • Surface better starting points (sites, contracts, budgets) • Reduce repeated data entry • Make progress and blockers visible to all parties At a high level, Phoenix supports the following flow: 1. Sponsor identifies potential sites 2. Sites are invited and complete feasibility 3. Required documents are populated, reviewed, and signed 4. Contracts are drafted, negotiated, and executed 5. Budgets are proposed, negotiated, and finalized 6. Site startup status is summarized and tracked Each step builds on the previous one. Data collected early is reused later to reduce effort and errors.

Ben Harrower

Police Transparency Project

Students will build an automated police transparency tool that transforms Natick police and arrest logs PDFs into a structured database that will enable journalists and the public to see trends and explore the data. They will create an interface that displays arrest logs as a heat map for Natick, as well as other towns, given time permits. If possible, students will enable newsrooms to link the database of incidents and arrests with select incident reports, which JO students will obtain through public records requests. Ultimately, students will develop pitches for news groups on trends to cover, and create public-facing data search tools and visualizations for community accountability. This system will serve as a replicable model for small community nonprofit newsrooms across the region to enhance police transparency and community oversight. The tool will address the current challenge where police and arrest logs are largely inscrutable and make it difficult for journalists and community members to track patterns, stay informed, and hold police accountable. By automating the data extraction and linking process, the system will enable continuous, systematic police transparency reporting that would otherwise require prohibitive manual effort for small newsrooms.

Bob Brown

Prison Overcrowding

Open Justice Lab

ML / Innovation · Spring 2026

Public Safety and Community Sentiment

Eric Gordon

Retrieval Augmented Generation

This is a continuation project building on substantial work completed in Fall 2025, when a Spark! team modernized Digital Commonwealth’s Retrieval-Augmented Generation (RAG) search pipeline. That work established a strong technical foundation, including a vector database, metadata-aware search, and a developer-friendly interface. In Spring 2026, the focus shifts from infrastructure to search quality, evaluation, and usability. Students will work on improving how well the system retrieves relevant materials, how performance is measured over time, and how results are presented to users. To keep the scope manageable, the project will focus on the BPL subset of the Digital Commonwealth collection. The ultimate goal is to support a natural-language search experience that returns results that are accurate, contextually relevant, and easy for non-technical users to understand. Project Goals 1. Incorporate Full-Text Document Content into Search • Extend the existing retrieval system to index and search the full text of documents, not just metadata fields • Experiment with combining document text embeddings and metadata signals (e.g., title, creator, date) to improve relevance and ranking 2. Track Inputs, Outputs, and Metrics in a Structured Way • Log queries, retrieved documents, model responses, and evaluation scores • Store results in structured, reusable formats (e.g., CSV, JSON, tables) to enable longitudinal analysis and future dashboarding 3. Build a Gold-Standard Evaluation Dataset • Create a curated set of representative natural-language queries with expected relevant documents • Use this dataset as a consistent benchmark for evaluating retrieval and generation quality 4. Implement an Evaluation and Testing Framework • Experiment with evaluation tools such as Ragas, ContextCheck, Phoenix, or similar frameworks • Establish baseline retrieval and generation metrics • Run evaluations iteratively throughout the semester to track improvements over time

Boston Public Library

Rio Grande Valley Broadband Project

The student team will support the coalition’s ongoing advocacy by analyzing gaps between federal broadband data and community-level realities, helping to create a community-informed broadband access map and set of visuals. The project’s core focus is to make visible the true digital divide in the Rio Grande Valley and empower young people to advocate for equitable digital futures. Students will engage in spatial analysis, qualitative research, and data storytelling, comparing the flawed FCC broadband map with more accurate, localized data, and using case studies or stories from the community to bring attention to broadband injustice. The resulting analysis will feed into public engagement campaigns and policy conversations around broadband infrastructure, affordability, and adoption.

Connect Humanity

Sewage and Algae Blooms

This project aims to create a potential warning system for Harmful Algae Blooms (HABs) by tracking multiple pollution pathways including Combined Sewer Overflows (CSOs), toxic runoff, and nonpoint source pollution — environmental issues particularly acute in New England. • We will collect and monitor federal, state, and local data describing when and where CSOs occur, while simultaneously tracking water quality chemical indicators that signal toxic runoff events. This integrated approach combines CSO discharge timing and precipitation-triggered water quality anomalies (indicative of potential toxic runoff) to create a comprehensive pollution event detection system. • Using satellite imagery acquired 24-72 hours following identified pollution events (CSOs, water quality runoff indicators, potentially heavy rainfall as well), we will monitor and detect HABs using the ABD (Algal Bloom Detection) repository, a production-ready open-source system that combines machine learning algorithms with satellite data analysis. The ABD system provides sophisticated anomaly detection capabilities specifically designed for HAB identification using multiple satellite sensors and validated machine learning approaches. • The system will correlate ground-measured chemical pollution indicators with satellite-detected bloom formations to improve prediction accuracy through ABD's integrated validation framework that combines One-Class Support Vector Machine, Random Forest, and Isolation Forest algorithms for robust detection. • We will eventually (in the spring of 2026) design an accessible app for New Englanders that displays CSO locations, common toxic runoff sites, and HAB formations with their pollution sources identified. Users will receive notifications when HABs exist in their area, with specific information about the likely pollution pathway (CSO vs. runoff vs. combined sources). The app will enable community members to contribute both photographic documentation and simple chemical test results to validate ABD's automated detection results. This project addresses the critical gap left by potential federal cuts to EPA monitoring systems, particularly the cyAN web app, while providing environmental justice communities with comprehensive water quality information across multiple pollution pathways.

TBD

Space-Time Data Explorer

The Space-Time Data Explorer aims to be an interactive data visualization tool designed to improve how users explore and interpret time-based and geospatial datasets published through the City of Cambridge Open Data Portal. Many datasets in the portal contain rich temporal and spatial dimensions, but are difficult to analyze using existing static tables or isolated visualizations. This project proposes a general-purpose, web-based exploration tool that allows users to visually examine how events, trends, or conditions change over time and across geographic space. The tool emphasizes intuitive interaction, enabling users to explore datasets through linked maps, timelines, and filters rather than requiring advanced technical skills. While designed around Cambridge data, the project is intentionally platform-agnostic so it can be adapted for other cities that use Socrata.

City of Cambridge

State Police Missing Data

Granite State News

UX / SWE · Spring 2026

Symporter

The project is to design and develop a web platform that connects BU undergraduates interested in pre-health research with faculty/PIs managing active labs. The platform will serve two primary user groups: • Students: Discover and apply to labs that fit their qualifications, interests, and career goals. • Faculty/PIs (Hiring Professionals) : Post lab opportunities, specify expectations, and identify qualified student applicants more efficiently. • Admin: Maintenance and oversight

Farah Rexha

ArtifexAI: Tech-focused Legislation Tracker / Predictor

Tech Policy Tracker

The Tech Policy Tracker aims to become a go-to, user-friendly platform for understanding the U.S. technology legislation landscape across both state and federal levels. While many existing policy tools cater to policymakers and researchers, this project prioritizes accessibility, clarity, and storytelling for everyday users For this semester, the project will focus on building a policy storytelling and visualization tool rather than a full end-to-end policy search engine. Using existing policy data provided by the client, students will explore ways to visualize tech policies and design and implement an interactive interface that allows users to explore, compare, and contextualize technology-related legislation. Users should be able to see which bills have passed, failed, are in progress, or were blocked, and understand how policies differ across states and over time. Some data visualization website are provided below in the "Relevant Documents” section as inspiration for the project. Students will work closely with the Shiran to refine user stories, define data structures, and design visual narratives that allow users to “enter” the policy landscape based on their interests, questions, or location. Emphasis will be placed on UX clarity, interactive visualization, and thoughtful framing of policy data, rather than policy analysis or lobbying.Ideal Output & Final Deliverables • A polished, interactive web-based dashboard that visualizes state and federal technology-related legislation • Map-based and/or comparative visualizations showing legislative activity by state, policy area, and bill status (e.g., passed, proposed, failed)

TikTok Professional Project

TBD

Unified Commerce Analytics

Superfeet currently relies on multiple disconnected dashboards and manual reporting processes to understand business performance across channels. As media investments have increased across platforms, the team needs a unified, data-driven way to measure how those investments impact sales, customer behavior, and long-term customer value. In addition, Superfeet has several years of historical purchase data that has not yet been fully leveraged to uncover consumer insights. This project will focus on building an integrated analytics and visualization dashboard that consolidates e-commerce sales data, advertising spend, and customer purchase history into a single, decision-ready view. Students will work with real-world data from 2023–2025 to analyze purchasing behavior, customer retention, and lifetime value, while also evaluating how consumer behavior changes before and after media campaigns are introduced. The project will emphasize both business-facing reporting and deeper consumer insight analysis, helping Superfeet move from reactive reporting to proactive, insight-driven decision-making.

Holly Ramer

Versetal Information Systems

The objective of this project is to build a secure, private AI chatbot that allows technical leadership to query support ticket data using natural language. The system must be able to perform both semantic search (e.g., “Find tickets related to international travel”) and trend analysis (e.g., “Summarize the top 3 recurring infrastructure issues in 2025”). With this project, there are three core constraints: • Privacy: No data can be sent to public LLM training sets (e.g., no public ChatGPT API). • Platform: Google Cloud Platform (GCP) or Open Source (self-hosted). • Interface: A chat-based UI (Python-based).

Versetal

Warning System for Private Detention Capacity Project

The goal of this project is to research, develop, and prototype a data - driven system that tracks and analyzes government contracting and procurement activity related to private detention services across the United States. Rather than focusing primarily on detention capacity, the project centers on following the flow of public funds – identifying which firms receive contracts, what services are being procured, when and where contracts are awarded, and how bidding activity signals future expansion. Private detention companies play a major role in immigration and criminal justice infrastructure, yet contracting processes, award timelines, and vendor relationships are often difficult for communities, journalists, and advocacy groups to monitor in real time. This project aims to increase transparency by building tools that surface early signals in procurement activity before physical expansion or policy implementation becomes visible. Students will analyze public data sources such as: • Government contracting and procurement databases • Corporate earnings and investor disclosures • Job postings and hiring trends By integrating these indicators, students will develop a prototype early-warning system that identifies emerging contracting activity, vendor positioning, and geographic areas of increased procurement attention.

Kaija Schilde:

WASH/Center on Forced Displacement (Amira Aker)

The client is collaborating on a UNICEF-supported initiative focused on understanding and monitoring infectious disease outbreaks in humanitarian settings, with particular attention to WASH (Water, Sanitation, and Hygiene) conditions and environmental factors that influence disease spread. Currently, outbreak and WASH-related data exist across fragmented formats including narrative reports, spreadsheets, embedded tables, and Power BI dashboards. There is no centralized or standardized database, and direct access to raw datasets or APIs is limited. Much of the available data must be manually extracted from reports or dashboards, creating a significant time burden and slowing analysis. Additionally, inconsistencies in reporting over time make it difficult to build coherent trend analyses or compare outbreaks across locations. In some cases, data governance constraints (e.g., ministry-controlled reporting) further restrict direct access. As a result, although substantial data exists, it cannot yet be efficiently used to answer critical operational questions about outbreak dynamics, case and death trends, or WASH-related risk factors. The primary goal of this project is to create a centralized master database that consolidates all available WASH-related and infectious disease outbreak data so the client can: • Understand what data currently exists • Identify which indicators are being tracked • Analyze cases and deaths over time • Enable future outbreak monitoring, comparison, and predictive analysis To accomplish this, students will focus on automating data extraction, cataloging indicators, and building a standardized dataset that replaces manual collection workflows and lays the foundation for future analytical work.

WASH

Housing Navigator Massachusetts

Website Engagement Analysis

Website Redesign

The Faculty of Computing and Data Sciences is a catalyst for education, research, and innovation in computing and data science, CDS connects a plethora of liberal arts, science, and professional disciplines with its foundational fields of computer science, computer engineering, mathematics, and statistics. The goal of this initiative is to elevate the CDS website through innovative design strategy and research, producing a prototype that feels truly outside of the box. The redesign will explore how best to communicate what makes CDS stand out, highlight the achievements of both faculty and students, and position the unit as a sophisticated, intelligent, and creative place. This effort will include a high-level redesign of the entire CDS website while also conducting a deep dive into Spark!’s integration within the site.

Molly Kate

911 Call Data Patterns

• This project focuses on developing an aggregator for 911 call data across selected U.S. cities. 911 call records are currently siloed in individual city data portals, often in inconsistent formats and lacking standardization. Our objective is to build a data pipeline that collects and harmonizes this data into a single, comparable dataset for analysis. • The project aims to uncover patterns in 911 calls for service for police (as opposed to emergency fire or emergency medical calls) — including who is calling, when, and why — to inform public safety alternatives and policy. It also contributes to ongoing research into how emergency calls correlate with crime, fatal police violence, and social determinants of health. • A parallel initiative on crime data, maintained by AH Datalytics via the RealTime Crime Index, will serve as a complementary resource. Some supporting code already exists to aid in navigation and data processing.

Andrew Zaharia

Data Science / Innovation · Fall 2025

Affordable Housing Applications

The Massachusetts housing market remains extremely tight, with affordability and accessibility posing persistent barriers, especially for low-income households and households of color. While CHAPA monitors nearly 3,000 permanently affordable homes across the Commonwealth, the application and resale processes raise questions about who is applying, where they are applying from, and what systemic barriers shape these patterns. Last semester’s team provided a broad descriptive analysis of CHAPA’s applicant datasets, producing detailed visualizations of age, race/ethnicity, income, assets, household composition, and marketing channels across applicants. Their work highlighted important findings, such as the impact of simplifying the application process (which increased diversity in applicants), differences in how households of color and white households applied across towns, and disparities in income and asset levels by race. (Please look at the final report from Summer work folder for more details) However, many key questions remain unanswered, particularly around applicant movement patterns, portfolio bias, and the effects of age-restricted housing. Additionally, CHAPA’s data represents only a slice of the affordable homeownership landscape (suburban, predominantly white communities), which limits the ability to generalize findings without partnerships from other monitoring agencies. The goal of this semester is to analyze and provide insights into these unexplored questions, with a focus on geographic trends, demographic representation, and access disparities. Primary goals being: • Understanding where applicants currently live compared to where they apply, the geographic radius of their search, and how these patterns differ between families, seniors, race, and other demographic groups. • Identifying whether marketing strategies, listing sources, or application complexity impact applicant diversity and equity of access. The project will leverage CHAPA’s recent application data (including four additional months of data with large lotteries) and may expand its scope by integrating datasets from other monitoring agents and municipalities, particularly Boston, whose affordable homes are managed by the city.

David Gasser

BU CISS: Andrea Beltran Lizarazo

Andrea Beltran Lizarazo: Colombian Extraditions to the US

The media often spotlight high-profile extraditions of wealthy and violent drug lords—mostly from Latin America, especially Mexico and occasionally Colombia—while overlooking a more common but less sensational reality: the majority of drug-related extraditions to the U.S. likely involve low-level offenders. Andrea’s dissertation uses Colombia as a case study to document and explain this broader trend. In past terms, Spark teams have successfully extracted data from more than 2,500 extradition cases spanning two decades, relying on court files and executive documents from official Colombian sources. This term, we will build on that work with two main missions. Each presents technical challenges that will allow the team to practice and expand core data science skills, while also contributing directly to Andrea’s research: 1. Corpus: PDFs to dataset: We have a collection of roughly 600 PDF documents, each about 40 pages long, which include short official communications of approximately one page each. The task is to extract these communications and organize them into a structured spreadsheet. The team will apply techniques in text mining, PDF parsing, and data structuring, and Andrea will provide textual patterns to define the start and end points for extraction. • Learning opportunity: working with large-scale text extraction from complex documents, building automated workflows for data cleaning and structuring. 2. Entity extraction with NLP/LLM tools: We also have an existing dataset that contains references to attorneys. The task is to process this dataset using NLP techniques, including an LLM agent, to extract entities such as attorney names and roles. The team will need to address challenges such as typos, inconsistent formatting, and linguistic variability. A preliminary attempt achieved an accuracy rate of about 66–67%. With improved inputs and refined methods, the mission is to raise that level of accuracy. • Learning opportunity: applying natural language processing to messy real-world data, experimenting with LLM pipelines, and refining extraction accuracy.

Anti-Displacement Tool for City of Louisville

The BU Initiative on Cities led development of an Anti-Displacement Assessment Tool for the City of Louisville. The ADAT provides a framework to evaluate the impact of new housing developments on local displacement, accounting for neighborhood specifics like rental prices and demographic factors. By linking development affordability to local housing conditions, the tool helps cities like Louisville create sustainable growth plans that protect vulnerable communities, particularly those at risk of gentrification. The ADAT is creative commons (open-source) and fully functional and is hosted online (here), currently utilizing R and shiny.io. The tool needs a modernized tech stack and user-friendly interface to allow for more scalability of the product. The goal is to redesign the tool for the city of Louisville with the hopes that it can be a model for other cities to embark on growth plans that elevate the community, ensuring affordable housing and preventing displacement while fostering inclusive and sustainable urban development.

Andre Comandon

Asad's Chatbot

This project will design and develop a web platform that empowers users to create, customize, and deploy their own chatbots without needing technical expertise. The platform will provide an interface where users can personalize their chatbot’s personality, design, and data connections, then generate a lightweight script that can be embedded on any website. The end goal is to enable individuals, organizations, or businesses to quickly build and integrate a branded conversational assistant into their online presence. The student team on this project will get exposure to developing web applications and AI applications, including architecture, tooling, and deployment considerations.

Asad Malik

Boston Climate Action Network

BCAN

Appending multiple building databases from the City of Boston, our team should understand the quality of the data, any missing data points, changes in reporting and emissions over time, and how these factors correlate with characteristics of the buildings, owners, and locations.

UX · Fall 2025

Boston Immigrant Resource Database

In partnership with BU Spark!, UniteBoston is developing the Boston Immigrant Resource Database (BIRD)—a digital platform designed to help immigrant-serving organizations find and share real-time, accurate information about available services. BIRD will feature a traffic-light availability dashboard, a searchable database, and a map-based interface focused on four critical service categories: legal services, housing, workforce development, and ESOL/citizenship classes. The DS488 UX team will enhance the provider-facing interface of the MVP and prioritize thoughtful user experience design that builds on existing development work. Strategy: • Prioritize the four key service areas for the MVP (e.g. Housing, Legal Services, ESOL, Workforce/jobs) • Focus on service providers as the main users to improve referrals. • Develop a dashboard that helps providers view and update service availability in real time. • Design for scale with a model that allows providers to self-update and potentially integrate automated reminders. • Partner with immigrant-facing organizations to ensure adoption, relevance, and long-term sustainability. Key Goals include: • Understanding user needs across immigrant-serving organizations through interviews and journey mapping • Designing an intuitive dashboard that allows service providers to browse, filter, and assess real-time availability of resources • Research update workflows that enable providers to confirm and revise their service status with minimal friction • Prototyping an interface, considering accessibility, mobile [a]usability, and multilingual readiness. See Overview section of BIRD for more details. Current Prototype & Development Progress: The current prototype, developed by Richard (a contracted developer), provides a foundational demo to explore the workflow and core features of BIRD. The intent of this is to test functionality and inform UX design work this semester. You can view Images of the prototype under the Client Meeting Notes tab. Key Prototype Features: Login • Support admin and provider roles • Admin accounts are invite only, providers can only edit their own services and resources Services Tab • Displays a list of services • Search tool which allows search by name, type, title, and description • Filters available for narrowing down by category • Service status options (traffic light): Open, Contact, and Closed Providers Tab • Search and filter providers by service type, keywords, language, and name Provider Details Page • Displays full organizational information and services offered Client Referral Form Page • Allows providers to refer other providers Alignment with DS488 UX Goals This prototype establishes the baseline functionality, but significant UX opportunities remain: • Refining the dashboard for clarity and accessibility. • Designing intuitive update workflows to reduce friction for providers. Prototype Access The demo is available at bird.softr.app/login [b](log in information to be provided by Parker and Richard) However, screenshots can be found under Client Meeting Notes.

Kelly Fassett

Boston Police Index / Shira Diner

Boston Police Index

Busing Project

Fifty years after the desegregation mandate for the Boston Public Schools, Councilor Weber would like to create a map tool to visualize the demographics of Boston Public School’s (BPS) facilities to understand the current state of school integration. Despite efforts to promote diversity through busing programs, significant segregation and disparities in school performance persist–there are vast differences in academic outcomes depending on the resources at different schools. Several Boston City Councilors have filed a hearing order addressing concerns about rising busing costs, long student commute times, and limited access to nearby schools. In response, our project will develop an interactive mapping tool that visualizes demographic patterns of school-age children across Boston’s neighborhoods, overlaid with the locations and grade levels of BPS schools. This tool will help evaluate the impact of current busing policies on school integration and equitable access to education. By analyzing enrollment trends, demographic shifts, and historical context, we aim to assess whether busing has met its original goals—or if new approaches are needed to better support Boston’s students.

Bonnie Deluaune :

Catering Leftovers

Massachusetts generates over 1 million tons of food waste each year, which is 25% of overall waste in the Commonwealth. At the same time, Massachusetts also has experienced surges in household food insecurity rising from 8.2% before the pandemic to 19.5% of households food insecure at its peak in Massachusetts. BU Dining is eager to develop a plan to respond to a change in Massachusetts regulations governing commercial food waste that requires facilities who generate more than one-half ton of food and other organic wastes per week to develop a composting program to eliminate waste. Catered events can generate a significant amount of wasted food. BU Dining is seeking a solution to give this (perfectly good) food away to students before resorting to composting. The challenge is that the food can only be given away within a set amount of time before it is subject to health and safety regulations. The BU Dining Leftover Food Website aims to solve the problem of redistributing leftover food from catering within these time constraints. This website is a fully functional, admin-facing platform that enables staff to efficiently log and post leftover food from catered events. Link to current web platform: https://bu-sustainability-leftovers.netlify.app/ Current core features: • Sign up and create profile • See FAQs • See events • Admin view by entering an admin token • Allow Admin to post events Following the initial launch, the web app’s functionality has been validated, and the team is now preparing to expand to the app store with the goal of making the platform accessible to the entire BU community.

Lawrence Alabaster

Computerized Mapping of VIsual Fields

The current approaches to visuospatial assessment in neuropsychology and behavioral neurology are analogue, paper-and-pencil type tests, that are far outmoded in the digital era. In contrast, retinal perimetry or visual field tests in ophthalmology (which maps deficits in ocular fields) have long been digitized – using computerized presentations and analysis. Visuospatial assessment, which looks at higher brain processes in spatial cognition, is in urgent need of a data-centric software platform for medical researchers and professionals to make it easier to (a) map out visuospatial dynamics, and (b) link patient performance data to brain imaging data. The first step in this process is to create a computational grid-based visuospatial assessment platform. The goal of this project is to create a grid-based visuospatial assessment platform for medical researchers and professionals. This platform will enable the mapping of visuospatial dynamics and the integration of patient performance data with brain imaging data. The app will consist of two screens: 1. A clinician’s screen to configure test settings and monitor patient performance 2. A patient’s screen to conduct various visuospatial tests. Currently implemented tests include: • Letter Test: Selecting letters on a grid • Color Change Test: Identifying changes in color on one half of an image • Flashing Test: Detecting a flashing half of an image Tasks Completed last Semester: • Patient Intake Form • Patient Database • Grid Color Change • Half of the House Color Change • Flashing Half of the House • Control and Navigation Panels

Dr. Vinoth Jagaroo

Data Liberation Project

MuckRock recently received a cache of records on slot machines located on U.S. military bases worldwide. This data comes from the Army Recreation Machine Program (ARMP), which currently operates 1,889 slot machines in 79 locations abroad and made $70.9 million from its slot machine operations during the 2024 fiscal year. The data currently exists as tabular information embedded in PDFs and has not been cleaned. The goal of this project is to extract the data from the PDFs, clean and standardize the fields, and build an initial exploratory experience that surfaces revenue and other key metrics by branch and region. This first pass is expected to generate follow‑up questions and further insights what this data can uncover. Cleaning and mapping this data could help explain which bases present the highest risk for military personnel with a gambling problem, what types of machines and games draw in the most money and why slot machines on American military bases have generated a steady increase of revenue since 2020.

Dillon Bergin

Division of Infectious Diseases: Strong Hearts Chagas Disease Project

Doubled Up Homelessness

Homelessness is a growing social problem with negative impacts on people’s health and life chances. Unfortunately, data on who experiences homelessness—and where—are imperfect, with scholarship consistently documenting the limits of traditional methods, from consistent undercounts to systematic exclusion of some forms of homelessness. One key methodological challenge is estimating the extent of doubled-up homelessness, defined as staying with friends or family temporarily because of housing loss or economic hardship (U.S. Department of Education, 2001). Estimates of doubling up can help communities understand the full spectrum of homelessness. Further, because most people who enter literal homelessness were last staying in a doubled-up situation, data on this form of housing insecurity can inform strategies for homelessness prevention. The purpose of this project is to pilot a dashboard or similar online resource that allows users to visualize and download data on doubled-up homelessness.

CISS Molly Richards

Facial Recognition for Stolen Artifacts

As part of her master’s thesis at Boston University, archaeology student Hallie Baker is developing a machine learning system to help identify looted Cambodian artifacts held in museum collections. Beginning in the 1960s during the country’s civil war and continuing into the early 2000s, Cambodia was heavily targeted for cultural looting. This looting was driven by Western museums and collectors, and today the Cambodian government is proactively pursuing repatriations from such institutions as the Metropolitan Museum of Art. Currently, researchers must manually search through thousands of archival photographs to find a match to a statue on display in a museum today—a slow and labor-intensive process. Yet associating these images is very important: the photos can raise red flags about the legality of an object or aid in the repatriation process by demonstrating proof of Cambodia’s ownership. Hallie is working to automate the matching of these images by building a searchable image database (currently containing over 600 images and 209 verified matches) and training a CNN model. Her project aims to eventually power a public-facing website where users can upload an image and receive potential matches. Her thesis lays out the need for this infrastructure, the technical steps required to build it, and future plans to expand the project to Indian and Nepali artifacts. The project is designed to be open-source and ultimately contribute to global heritage preservation.

Hallie Baker

Harvard Allston Project

This project centers on identifying elderly residents in the Allston-Brighton area who may be eligible for newly planned affordable senior housing units. The goal is to help ABCDC build a foundation for equitable outreach and data-informed planning by methodically identifying and profiling older adults in North Allston and beyond who may benefit from housing opportunities. Collectively, the team will combine data analysis, mapping, and system design to support outreach to aging residents, assess housing readiness, and forecast the impact of new affordable housing developments. “This project focuses on identifying and understanding the elderly population in the Allston-Brighton area, with an emphasis on informing future affordable senior housing initiatives. The team will work with ABCDC to develop a clear, data-driven profile of older residents, including where they live, their housing situations, and potential needs for relocation or supportive services. Using a combination of public records, census data, and ABCDC’s own housing information, the project will map concentrations of elderly residents, assess tenure and ownership patterns, and identify key indicators such as income, housing condition, and accessibility needs. This analysis will help ABCDC strategically plan outreach, prioritize neighborhoods with the greatest needs, and anticipate how new affordable senior housing could affect both individual residents and broader neighborhood housing dynamics. The deliverables will include a cleaned and structured dataset, geographic visualizations of elderly resident distribution, and a narrative report summarizing findings and recommendations for outreach and planning. The project will give ABCDC a stronger foundation for equitable, targeted engagement with the elderly community in Allston-Brighton.”

John Wood

Harvard Berkman Center's Applied Social Media Lab

Kalie Mayberry

UX · Fall 2025

HeatWise Citizen Science

Building upon the Spring 2025 teacher-dashboard prototype, the DS488 team will refine the Heatwise Citizen Science app, specifically targeting high school students. The app's primary functions will focus on structured environmental data collection and the management of an accessible data repository. Boston Public School educators will seamlessly assign, supervise, and organize these environmental data activities, while students engage in meaningful field data collection. Supplementary instruction provided by teachers will add educational depth and context beyond the app’s functionality. This dual-persona platform will align closely with the curriculum standards set by Boston Green Academy and Boston Public Schools. Current Prototype Status The current prototype is not yet at an MVP level. The main objective remains educational data collection as it is explicitly not intended as a homework platform, nor should it emulate competitors previously researched. The platform is anticipated to function primarily as a web app, enabling compatibility across multiple devices. Key Features Data Collection and Repository: Clearly defined data collection tasks aligned with ecological science objectives (soil composition, ambient temperature, etc.) Minimal Gamification: Simple progress tracking and basic rewards (badges, points) to motivate/track consistent data collection. Gamification strictly supportive of data accuracy and educational outcomes, avoiding distraction Teacher and Student Interfaces • Teacher • Invite, edit, and manage student accounts • Initiate and oversee data collection activities • View student progress and data submissions • Edit and manage datasets • Basic customization of data visualizations • Download data into csv files • Student • Streamlined login • Clear, intuitive data collection workflow • Brief context blurbs about data collection tasks • Progress indicators • Minimal reflections on data collection for teacher review • Basic achievement badges/statistics reflecting their data collection. Defining the data model: finalize what variables are captured onsite (e.g., surface temp, canopy cover photos, soil pH) and how they flow to REDCap or CSV for researchers.

BU Innovate

Jonathan Cohn, Hewon Hwang

Home Rule Petition Project

This project will analyze the lifecycle of home rule petitions as they move through the Massachusetts legislature. Municipalities across the state frequently submit these petitions to seek exemptions, special powers, or local policy flexibility. While some succeed, many fail – often requiring repeated filings year after year. By mapping this process and studying outcomes, the project aims to uncover patterns of approval, failure, and repetition that shape local policymaking. This analysis will provide policymakers, municipalities, and researchers with a clearer understanding of how the legislature treats home rule petitions. By identifying structural bottlenecks, success factors, and fatigue effects, the project can inform both local strategy (for towns and cities filing petitions) and state-level reform discussions about how to streamline the process.

House Roll Call Tracker

The Massachusetts state legislature's website presents significant accessibility challenges when it comes to tracking and analyzing roll call votes. Currently, all votes are published as individual PDFs and usually unlinked from the bills they actually concern, making it difficult for users to find relevant legislative records and assess lawmaker voting behavior. Furthermore, the number of recorded roll call votes has been decreasing each session, reducing transparency. This project aims to develop a data pipeline and interactive visualizations to streamline legislative vote tracking and analysis.

Scotia Hille

ML · Fall 2025

Identifying Children's "Stranger Danger" Behaviors

One of the most common mental health problems among children are anxiety disorders, with many children maintaining their anxiety well into adulthood. Anxiety impacts children’s ability to make friends, develop proper social skills, and participate in class which leads to poor educational outcomes. Therefore, it is important that researchers find ways to identify which children are at most risk for developing anxiety as early as possible to get them the treatments necessary. One of the most common predictors of anxiety is behavioral inhibition, which is an early personality trait that is characterized by fear and withdrawal to new situations and people. Being able to identify these behaviors helps us identify which kids should receive extra support to reduce their chances of developing anxiety later in life. Traditionally, researchers and therapists will watch how each child reacts to new situations and identify a variety of behaviors that are classified as risk factors for anxiety. While this type of behavioral coding is the most common and accurate way to assess risk behaviors among children, this is very time consuming to watch hours of videos and requires many hours of specialized training to make sure all behaviors are identified in the same manner across different coders. Recent advancements in machine learning have created the ability to analyze human movement and behavior using computer vision, however few researchers in the field of psychology have attempted using computer vision to digitize their traditional behavioral coding schemes. This continuation project builds on work from Spring 2025, where the student team successfully developed a web-based tool for analyzing video footage of children in "Stranger Danger" scenarios. The system labeled children’s proximity to others in the room and estimated fear and freezing behaviors using pose estimation and facial recognition models. The goal of the Fall 2025 project is to further improve the robustness, scalability, and accuracy of the existing system. Key areas of focus include reducing frame detection failures, increasing prediction precision, adding behavioral and audio event detection capabilities, and expanding these features to a “scary robot” scenario directly following the "stranger danger” scenario. Proposed extensions include: • Enhanced person detection in relation to a static location (parent chair, robot) to improve accuracy of proximity • Behavior recognition possibly by utilizing pos detection: approach, withdrawal, gaze shifts, speaking, crying • Audio-event synchronization: detect when the stranger speaks, child responds, toy is offered/taken, when experiment is started More detailed labels for the data are in the experiment and please look at the previous work under the background reading section to

CISS Kathy Sem

ML · Fall 2025

Image Classification Pipeline

WLFC

Job Search Project

This project will explore and define what types of jobs in the humanities and STEM (science, technology, engineering, arts, and mathematics) can be classified as Public Interest Technology (PIT) careers. While PIT has often been associated primarily with computer science and engineering, there is growing recognition that technology intersects with a much wider range of disciplines. Students and professionals in the humanities and arts contribute critical perspectives on ethics, accessibility, communication, design, and community impact – all of which are essential for ensuring technology serves the public good. The project will begin with research to establish a framework for identifying which jobs across humanities and STEM fields can be considered PIT roles. This will involve reviewing existing PIT definitions, analyzing job postings and descriptions, and consulting with faculty, practitioners, and employers. Once identified, these roles will be organized and categorized into an accessible, student-facing format that highlights career pathways into public interest technology beyond traditional technical roles.

Carolina Rossini

Data Science / ML · Fall 2025

Language Predictor Analysis

This project focuses on predicting a voter’s primary language using voter record data containing 5.7 million rows, including around 160,000 labeled records and over 400 features, with the goal of improving multilingual voter outreach across Massachusetts. Last semester, students trained and deployed machine learning models on over 13,000 labeled records to classify languages such as Haitian Creole, Cantonese, Khmer, and Mandarin. Students last semester took two approaches: • Name-based modeling – Generated language likelihood scores for individual voters based on their names. • Precinct-based modeling – Language probabilities were inferred from precinct-level voting distributions, capturing geographic patterns of language use The outputs were combined into a dataset that appended probability scores for each target language, enabling outreach teams to identify which communities may benefit from multilingual materials. The goal of this project is to improve predictions for languages such as Cantonese and Khmer, as well as to combine the name- and precinct-based models into a single ensemble model. While the primary scope this semester is to build upon and refine the previous work, students are encouraged to explore and experiment with alternative approaches if they arise, and to discuss potential improvements with the client. It is recommended that students have some knowledge of regression models (using the scikit-learn package or similar), familiarity with feature engineering, and experience with the pandas package in Python.

Massachusetts Voter Table

Media Coverage Analysis

CAIR has hypothesized that there has been a decline and shift in local media coverage related to Muslim and Palestinian communities in Massachusetts since October 2023, particularly in The Boston Globe and MassLive. Prior to this date, these outlets often covered community reports and letters that elevated Muslim voices. After October 7, 2023, CAIR believes there has been reduced coverage, less representation of Muslim authorship, and framing that disproportionately favored Israel. This semester, the student team will explore and test whether these hypotheses hold true. The work will begin with an elementary keyword-based analysis of relevant news coverage, followed by the application of NLP techniques such as entity tracking to identify who the subject, names, organization, locations, and date. Students are also encouraged to experiment with methods to quantify whether recent articles demonstrate associations with Islamophobia-related topics, even if only at a prototype or exploratory level. Nationally, CAIR and ISPU have developed tools such as the Islamophobia Index and reports like Hijacked by Hate to capture themes of Islamophobia (e.g., portraying Muslims as violent, misogynistic, or anti-Western; exclusionary editorial decisions; or unequal treatment). However, these tools are theme-driven, not keyword-driven, which poses challenges for automation, they can serve as reference points to help guide idea generation and inform exploratory analysis in this project. For data collection and exploration, the team will use two approaches: • The Boston Globe (via ProQuest): Access will be through ProQuest, a subscription-based research database that provides archival and full-text access to newspapers, journals, and magazines. A key limitation is that ProQuest does not allow articles to be exported outside its environment. Instead, students must use ProQuest’s built-in Jupyter Notebook console to run text-mining and data analysis experiments. Only results, outputs, and summary findings can be exported—not the full text of the articles. • MassLive (via Web Scraping): Since MassLive is not indexed in ProQuest, its content will need to be collected through web scraping. Fortunately, MassLive maintains a sitemap that organizes articles by category and date, which should make scraping and filtering content relatively straightforward.

Tahirah Amatul-Wadud

Medical Debt

This project will analyze Massachusetts court data from MassCourtsPlus to identify and investigate lawsuits related to medical debt. The primary objective is to uncover the true extent—and minimum credible estimate—of medical debt lawsuits filed against individuals across the state. Using a combination of keyword filtering, plaintiff classification, and reverse-engineering of public methodologies used by the Massachusetts Trial Court, the team will focus on identifying key creditors (e.g., CareCredit/Synchrony, Comenity/Bread Financial, Wells Fargo) and debt buyers actively pursuing such cases. Special attention will be paid to entities filing large volumes of lawsuits and the diverse legal forms medical debt may take (e.g., dental, vision, ambulance, elective procedures). The analysis will directly support GBH’s investigative reporting by highlighting patterns in debt collection, potential disparities across courts, and the broader impact of these practices on individuals’ financial stability, health, and well-being. Over the summer, a team collected and standardized debt collection cases from MassCourts.org, cleaning plaintiff names for accuracy. From this dataset, we isolated and analyzed key subsets, including MSPCA Angell, Synchrony, and ambulance service cases. We also consolidated and analyzed the Division of Banks Debt Collector Annual Reports (2022–2024), preserving survey details to produce debt type classifications and trend analysis. This semester Jenifer would like to better understand how much of the debt lawsuit landscape can be confidently attributed to medical-related claims and look into the ambulance service cases and undercover how they are handled in Massachusetts court. I would suggest two focus areas: (1) Ambulances (2) Vets/Hospitals (including MSPCA - Mass Society for Prevention of Cruelty to Animals) Last 10 years only- 2014 until current

GBH

Granite State News Collaborative

NH Police Stops

This project is years in the making, so it is important to understand its somewhat sordid history. Back in 2023, a team began this project to build on Granite State’s 2022 investigative series about NH's Mobile Enforcement Team (MET), which found this drug-interdiction unit was using minor traffic violations as pretexts to conduct searches that disproportionately impacted Black and minority drivers. The initial Justice Media teams thought they found disparities, but another team in 2024 figured out the data was severely flawed and missing race in many records. With a lot of reporting and data analysis, they figured out the New Hampshire State Police had stopped collecting race data altogether! Reporting and writing this story about how they aren’t collecting this data is one of the JO teammate’s tasks this semester. While the spring 2025 team was reporting on the missing race data, we came to legally possess data that includes additional fields including names of the people who were stopped. With this new data, we want to try to assign race to the names with enough accuracy to determine if there is still a racial disparity in stops. This project consists of two stories: Story #1 Missing race data & impacts Story #2: Racial disparity in stops with new data The team will investigate: • Whether racial disparities persist in available data • Geographic and temporal patterns in traffic stops • The relationship between initial stop reasons (e.g., speeding) and subsequent searches/arrests

ML · Fall 2025

Registry Transcription

The Hampden County Registry of Deeds holds a vast collection of historical records dating back to the early 1600s. With approximately 100,000 documents spread across 600 books (each containing around 1,000 documents), these records provide invaluable insights into property ownership, land transactions, and historical boundaries. However, due to the handwritten cursive nature of the documents, inconsistent document delineation, and the lack of a comprehensive indexing system, accessing specific records remains a challenge.

Laura Marino

SWE / UX · Fall 2025

Social Justice Mobile App

The goal of this project is to develop a tool that will allow Dr. Starks to engage with clients through a comprehensive course platform, streamlining his mission of empowering individuals and organizations to navigate complex social justice challenges. Additionally, this platform should act as a social justice-focused community hub, fostering equitable and inclusive spaces where leaders can host community spaces and their members can voice their opinions, have important conversations, and facilitate their own journey of a more diverse, inclusive, and equitable organization or community. For the Spring 2025 semester, this project will focus on the following priority feature areas: community spaces, user onboarding and monetization strategies, and Dr. Stark’s personal coaching.

BCS & Associates

Boston Attorney General Office

State-Wide Resource List for Trafficking Survivors

Styrofoam Ban Project

Boston is trying to ban polystyrene (Styrofoam) foodware in establishments (retail and food). Similar bans or restrictions already exist in over 50 municipalities across Massachusetts. Pholystrene is a pollutant as it does not biodegrade, contributes to microplastics, and is excluded from Boston’s acceptable recyclables, and its main chemical styrene is possibly carcinogenic, posing risks to respiratory and neurological health. At the same time, enforcement and compliance require balancing environmental justice with economic equity. Small businesses need time, funding, and technical assistance to transition to alternatives. This project will support Councilor Zapata’s office in evaluating the impacts and considerations of a Styrofoam ban, specifically with a cost based focus regarding small local businesses.

Ellie Sanchez

Thurmond Amendment Project

Yusuf leads a national campaign to pass the Fair Future Act, which would repeal the Thurmond Amendment and restore fair housing protections for individuals with a drug distribution conviction. Yusuf wants students to analyze drug distribution disposition data to understand the scope and detrimental impact of the Thurmond Amendment. He will use that data in conversations with advocates and federal policymakers to help pass the amendment. This project is a multi-state analysis, focusing on data derived from open records requests in Arkansas, Washington, and Wisconsin.

Thurmond Amendment

Unified 5K

Unified5K is an application designed to provide information and personal stats on 5k races put on by the organization. It aims to provide easy access to race information for all participants and spectators by utilizing race cards with start times and location, badges and personal stats for racers, as well as donation and sponsor information. With racers and spectators in mind, it ensures information is easily accessible to all while keeping accessibility features in mind for the intended audience. Key Features • Tabbed navigation: Home, Media, Resources, Donation, Profile ✅ • Race Results & Tracking (Race Cards): supports searching/filtering ✅ • Race details: Descriptor, image carousel, and donation progress ✅ • Donation Page - Sponsor tiers: Collapsible sections with summaries + “Learn more” links✅ • Sponsor/Vendor inquiry: In-app modal (mailto) to contact AdaptX ✅ • User Profiles: Includes race history and stats Check In progress • Social Feed: blog posts In progress • Account login/creation with email and password ✅

Brendan Aylward

Vaccine Equity

In Massachusetts, the most effective vaccination efforts for LatinX and Black communities were led and organized by grassroots organizations. Groups like La Colaborativa worked to create a vaccination site with East Boston Neighborhood Health Center to provide vaccines for the Chelsea community. Yet, other entities claimed credit for the high vaccination rate while grassroots organizations worked for these communities. In Fall 2022, students compared the vaccination rates of two cities, one that doesn’t have a strong grassroots organization and one that does: Revere and Chelsea. Large hospitals have clinics in both cities that provide non-reimbursed services to maintain their not-for-profit tax status (it is their obligation under tax law). The goal was to compare the vaccination rates in these cities - if the hospital clinics were providing the critical services, then the vaccine rate in Revere should be similar to that in Chelsea. In Fall 2023, students explored New Bedford and Wellesley. In fall 2024, students generated an interactive map that enabled comparison of cumulative vaccination rates by certain demographics (age and race/ethnicity groups) in different cities. This semester, Dr. Koehler would like us to continue the research done in previous semesters and expand our analysis to include additional cities similar to Chelsea but outside the scope of La Collaborativa. The confirmed cities are Lawrence, Holyoke, Springfield, and Lynn, with a broadened scope to include other comparable cities as well as some dissimilar ones with varying income levels and demographic profiles.. We will analyze COVID-19 death rates by year (2020, 2021, 2022) across these cities and compare them to Chelsea and wealthier towns. The project will also focus on fine-tuning visualizations to incorporate meaningful comparators, such as statewide averages for vaccination rates and other relevant metrics, to provide a more comprehensive context for the findings. The ultimate goal is to better understand the role of community-based organizations in improving public health outcomes and to advocate for their increased attention and funding.

Website Analytics

Alexa DeRosa

UX · Fall 2025

Website Redesign

Blackfacts.com, the first and only online encyclopedia of black facts and history, offers original videos, sourced articles, widgets for educators websites, and a plethora of knowledge. Created in 1997 by Ken and Dale, they continue to develop the site and bring black history to schools, school districts, and anyone looking to gain more knowledge. The goal of the BlackFacts and Spark! Partnership is to elevate the current website through a full redesign. The website holds a very large body of information, and lots of attention is needed to ensure this important information is digestible and accessible to the masses. Research has been done previously, so it is vital to build upon the past work to ensure no progress is lost.

Ken Granderson

· Summer 2025

Doubled-Up Estimate Coding

This project aims to improve how we measure “doubled-up homelessness,” which is when people temporarily stay with family or friends because they don’t have a stable place to live. Dr. Richard worked with community partners to use public American Community Survey (ACS) (Census Bureau) data to estimate doubling up. However, the estimates can have low levels of reliability due to the small sample size of the public-use ACS files. To address this limitation, Dr. Richard wants to apply to use the larger, restricted-use datasets, but needs help creating variables (described more below) not directly provided within this dataset (but that can be created using source variables). Broadly, the goal of the project is to develop the code needed to get more precise estimates of how many people are affected by this form of homelessness.

BU CISS: Molly Richards

AdaptX Athletic Guide Pairing Application

The goal of this project is to create an app that will assist those with visual and/or physical impairments by connecting them to the nearest “guide” athlete around them to compete in races and other athletic events. Users with visual and physical impairments should be able to register, create profiles, and connect with nearby guides or coaches. The app should be compatible with screen readers, voice commands, and other assistive technologies. Integration with Strava, NikeRun Club, and potentially other fitness tracking apps to allow users to track their progress and connect with a greater fitness community

Autonomous Mechanics Challenge

SWE · Spring 2025

Autonomous Mechanics Challenge

This project will turn a robotics lab into an online tool anyone can use from anywhere in the world. Users will log in securely, send in tasks for robots to complete (specifically 3D printing or testing parts), and track how their projects perform using visual tools like leaderboards. The platform should be easy to use while ensuring safety and smooth operation, creating a resource for researchers, and students.

Award Analysis

Academic Analytics is a platform that allows university leaders and faculty to benchmark their research output compared to peers. One part of their offering allows the benchmarking of prizes and awards. The client is interested in finding a way to predict faculty’s likelihood of winning an award based on the current set of awards that they have won.

BU Strategy & Innovation

Boston Climate Action Network

BCAN / Urban Emissions

Policies similar to BERDO exist in other major cities, including Chicago, New York, San Francisco, and Seattle, each tailored to their unique environmental, economic, and urban landscapes. This project aims to leverage data from these cities to analyze the effectiveness of building emissions policies, uncover trends in energy and carbon use, and identify opportunities for acceleration toward net-zero goals. Key areas of focus include: • Policy Comparison: How do emissions reduction policies in Boston, Chicago, New York, San Francisco, and Seattle differ, and what lessons can be learned from their successes and challenges? • Performance Evaluation: Which cities, building types, or sectors are leading or lagging in emissions reductions, and why? • Trends and Insights: What patterns emerge in building performance, and how do factors like size, age, or function influence emissions levels? • Predictive Analysis: What interventions or strategies could help accelerate progress in cities or building categories falling behind?

Behavior of Trees

Today, Season Watch uses a reference database to flag for errors in citizen submissions (data collected by the citizens). However, due to rapid changes in the climate, trees are going through their seasonal phases earlier or later than anticipated, so citizens’ inputs may not actually be inaccurate. Season Watch would like us to assess and understand the shifts of seasonal behavior of common trees by comparing citizen submitted tree data against the reference database. This is achieved by comparing citizen-submitted observations of tree phenology (such as leafing, flowering, and fruiting times) with a historical reference database. The goal is to identify discrepancies, infer actual shifts over a decade, and update the reference database based on emerging patterns to improve its accuracy and reliability.

Season Watch

Boston Medical Center: Neuro Assessment Website

Patients that undergo certain cellular therapies are occasionally at risk of developing dangerous brain toxicity as a side effect of their treatment. In order to test for brain toxicity, patients will undergo a neurological assessment. However, this assessment is currently only available in English and oftentimes patients are non-english speakers. Therefore, the goal of this project is to create a website-based replication of this assessment that will translate the assessment into different languages for non-english speakers.

Boston Medical Center

Catch Up Clubs

The goal of this project is to further analyze what enables students to progress through school and improve learner retention, aligned with the traditional school year. This project should identify predictors for success and failure, such as course drops, stagnation at certain reading levels, learner retention, attendance, disability status, socio-emotional level, and any other influential variables. This project should integrate predictive analysis into dashboards and advance their current analytics.

Save the Children

Children's Health Equity

The Hospital's Office of Government Relations hypothesis is that the Commonwealth is underinvesting in public health approaches to supporting children’s health and well-being and children’s primary and preventative health care services relative to adult health. Relative to supporting the health and well-being of children, the state’s health care and public health dollars are more heavily invested downstream to treat illness and disease, and in adult health care services. Massachusetts children living in low-income communities, communities of color, and rural areas receive less healthc spending and less access to high-quality healthcare services. Relative to similarly affluent states or countries, Massachusetts spends comparatively less on pediatric healthcare services and public health approaches to support the health and well-being of children. Students will analyze state government and health care spending (with a special focus on MassHealth / Medicaid) on children’s health, pediatric care, and well-being to identify underinvestment and gaps in spending on children and compare spending levels to national spending, similar states, or affluent countries.

ConnectEd

The goal of this project is to identify relationships between student information (mental health, sports, clubs, connections, demographics, etc) and their sense of belonging in the school they attend in St. Helens as a mechanism for youth suicide prevention. For example, one conclusion can be that students who are on the wrestling team generally have lower feelings of belonging because they have fewer connections. The survey provides information/data on individual students and staff as well as information/data on their connections. We want to analyze both. The data used for this analysis will be two rounds of survey results and exports from a social systems app called Kumu. Students are asked to name their 5 closest friends and share how they know them, how much time they spend with them, what sports and clubs they participate in, and much more. Additional Project Information

ConnectEd

FIO and Arrest Analization

To help OPAT’s mission of police accountability and transparency, students will analyze 3 datasets: one on Field Interrogation and Observation (FIO), one on arrests and Use of Force, which contains information about the person that got arrested or/and any police encounters such as their ethnicity and type of interaction/force. The students will create interactive dashboards to visualize BPD FIOs and arrests by race, year, location, and more to help answer the questions below (Look at base questions).

OPAT

Housing and Voting

• This collaborative investigation will analyze the relationship between ethnicity and voting patterns in New Bedford from 2020-2024. The secondary priority is to analyze the connection between housing status (renters vs. homeowners) and voting patterns. Using voter registration data, property records, and election results, the project will examine how housing status correlates with political participation across municipal, state, and federal elections. Through data visualization and narrative reporting, this project will help understand whether and how housing status influences civic engagement in New Bedford's diverse communities.

New Bedford Light

Initiative on Cities: Anti-Displacement Tool

The BU Initiative on Cities led development of an Anti-Displacement Tool for the City of Louisville. The ADT provides a framework to evaluate the impact of new housing developments on local displacement, accounting for neighborhood specifics like rental prices and demographic factors. By linking development affordability to local housing conditions, the tool helps cities like Louisville create sustainable growth plans that protect vulnerable communities, particularly those at risk of gentrification. The ADT is creative commons (open-source) and fully functional and is hosted online (here), currently utilizing R and shiny.io. The tool needs a modernized tech stack and user-friendly interface to allow for more scalability of the product. The goal is to redesign the tool for the city of Louisville with the hopes that it can be a model for other cities to embark on growth plans that elevate the community, ensuring affordable housing and preventing displacement while fostering inclusive and sustainable urban development.

Initiative on Cities

Kentucky NextGen Justice: Innocence Project

This project focuses on designing an online legal services toolkit, LyraLegal, aimed at addressing systemic challenges in the legal system through leveraging AI. This tool will help small law firms, nonprofits, and underserved communities manage legal issues more efficiently and provide people with the resources they need to legally advocate for themselves. Given that millions of people file legal cases without legal representation each year and 86% of individuals cannot afford legal services due to high attorney costs, LyraLegal seeks to bridge this gap by offering accessible and affordable legal resources. The platform will also simplify complex legal processes, such as filing for divorce, starting a business, or applying for legal aid, ensuring that individuals can navigate the legal system with greater confidence and efficiency. The product currently has a working backend and LLM, and now needs a user-friendly interface to help make this tool accessible to all.

Kentucky NextGen Justice

NU LawLab: MAPLE 3.0

This project aims to improve the design and user experience of MAPLE, ensuring it effectively serves a wide range of users—from experienced political leaders to individuals with an interest in law and policy. This semester, you will focus on refining MAPLE and designing its 3.0 version, with an emphasis on improving visual design, navigation, and content organization to enhance usability and accessibility. A key component of this project will be a re-skin of the website, aligning it with updated design guidelines to create a more intuitive and visually cohesive experience. Particular attention will be given to the Bill Details page, which will be restructured into a tabbed system to improve information hierarchy and user flow. Additionally, you will work on designing components that populate these tabs, such as lobbying disclosures, legislator voting records, and hearing transcripts, ensuring that complex legislative data is presented in a clear and digestible manner for a wide range of users.

NU LawLab

Predictive Model Based on Homelessness

Dr. Byrne and Dr. Richard are planning to put together a peer-reviewed article that expands on an existing body of research that examines structural determinants of homelessness across roughly 400 communities that receive federal homeless assistance funding from the U.S. Department of Urban Development. These communities, which are geographical units known as “Continuums of Care” (coc) (see Glossary for more information) have to conduct counts of homelessness on an annual basis. Prior studies have used these data to examine the association between community-level factors such as rent levels, demographic and economic conditions and rates of homelessness. By analyzing and using 2007-2023 data, the goal of this project is to move beyond simply identifying associations and instead develop a predictive model of homelessness at the community level using these same factors. While some studies have been done to predict homelessness at the individual level based on factors like income and location, this project is unique in its focus on the community level. This project will aggregate data from the Department of Housing and Urban Development, along with publicly available data such as rent rates (see the Dataset section for more information), and group them by Continuum of Care (CoC) unit. Dr. Byrne and Dr. Richard have a strong understanding of the predictors needed to build the model. This should be discussed further during the initial client meetings to gain guidance on how to structure the dataset. Following this, regression models or other predictive machine learning models should be developed to predict the number or rate of homelessness by CoC. It is recommended that students have some knowledge of regression models (using the scikit-learn package or similar), familiarity with feature engineering, and experience with the pandas package in Python.

BU CISS: Tom Byrne

Second Nature Project

Many colleges and universities have a goal to reach carbon neutrality. Some institutions are better equipped to meet their sustainability goals more aggressively than others. Consider how an institution that mainly serves commuter students may have a harder time reaching carbon neutrality than one that houses students on campus. Your goal is to identify factors, such as the difference between residential and commuter students, that may help or impede an institution’s ability to meet its sustainability goals. Given these factors, your team will identify which institutions and types of institutions have been best able to meet their goals and make recommendations as to how other institutions might better be able to reach theirs.

Second Nature

Sports and Academics Predictor

BU Athletics is interested in understanding how well a student will perform academically during their time at Boston University based on the student's high school GPA, geography, demographics, language proficiency for international students, what they choose to study, and standardized exam score. This will help BU Athletics understand 1) how they can better support future student athletes and 2) to assist our coaches in the recruiting process to understand if/where a student may struggle or thrive as a part of our community. The goal is to identify how we can better support future student-athletes by finding trends in areas where students may struggle or excel within the Boston University community. You will be provided with data from students who have already graduated, including information such as race, college, high school GPA, hometown, and semester GPA. This data needs to be analyzed to identify trends in the success or struggles of athletes. If time permits, students can also explore creating a basic regression model to make predictions based on the trained dataset. **Please note you will not be measuring athletic performance related to admissions data and instead will focus on academic performance.**

BU Athletics

Data Science / Data Visualization · Spring 2025

Team A Appropriations Equity Data Science Project & Equity in Federal Budget Earmarking Processes FY24-25

The practice of congressionally directed spending (CDS), previously called “earmarks” describes a mechanism through which members of congress send discretionary spending directly to their district or state for a specific project such as funding an infrastructure project, an organization, or other local initiatives. FY22 was the first year of CDS funding after a decade without earmarks. This project seeks to create a repeatable process for analyzing the process of evaluating CDS requests through an equity lens. Building on the Work of Spring 2024 & Fall 2024 Projects: The Spring 2025 project will be an expansion of previous semesters’ work on a data visualization dashboard to help the Senator Markey office better understand where federal funding is being distributed across Massachusetts by merging the existing dashboard with 3 new datasets; • FY22 to FY24 CDS Awards for BU Spark • Updated Invest.gov data 12.18.2024 • Reworked Regional and Statewide Grants We will use CDS requests and earmarks interchangeably in this document. This project will follow requests through the following steps: Earmark/ CDS requests from Senator Markey’s constituents → Approved CDS requests from Senator Markey’s office put forward as sponsored requests to various committees → Committee approvals → CDS distributions to Senator Markey’s constituents The project will seek to apply an equity lens to this process and understand who is applying for earmarks and who they will benefit as well as who is NOT applying for CDS, i.e. who is not benefitting from this opportunity as well as who is getting approved and whether there are disparities at this stage of the process. We are specifically looking at equity through the lens of race

Senator Ed Markey's Office

Senator Ed Markey's Office

Team B Competitive Discretionary Grant Allocation Equity Analysis

Each year, federal funding in the form of competitive discretionary grants is distributed to communities across Massachusetts for various projects and purposes. The goal of this project is to examine the equity of this allocation process on a local level, especially amongst BIPOC, low-income, and underserved communities. By creating a data analysis, repeatable process for metric reporting, an interactive site, and more, we hope to help Senator Markey’s office determine which parts of MA continue to be underserved in the allocation process.

City Councilor Benjamin Weber: Data / Constituent Engagement

Tree Canopy Coverage Analysis Project

This project aims to explore the distribution of tree canopy coverage in Boston to assess whether it is possible to distinguish between city-owned and privately-owned trees. The focus is on analyzing tree canopy coverage across different land use categories to understand disparities between trees in public spaces (e.g., parks, streets, open spaces) and those on residential properties. The primary goal is to experiment with available data to uncover patterns in urban tree canopy distribution. While the analysis may not fully determine the ownership status of all tree canopy, it is designed to provide valuable insights into how much of the canopy is attributable to city-managed versus private parcels. This experimental approach ensures flexibility and sets the expectation that partial insights can still advance our understanding of urban tree equity. The dataset includes high-resolution land cover data, LiDAR-derived tree canopy layers, and parcel-level land use data sourced from the 2014-2019 Boston Tree Canopy Change Assessment. Students will work with CSV files containing detailed metrics, such as existing canopy coverage, potential canopy areas, and changes over time. Using Python or similar data analysis tools, students will calculate metrics like canopy coverage percentages within public and residential parcels, aiming to identify discernible patterns. By treating this as a discovery-driven experiment, the project emphasizes the learning process and incremental findings. Any insights gained—whether complete or partial—will contribute to the understanding of urban canopy equity, supporting efforts to enhance urban forestry strategies, mitigate urban heat islands, and promote equitable access to the benefits of tree canopy coverage in Boston. This project ultimately aligns with broader goals of environmental justice and sustainable urban development. Preferred Client Meeting Time: Thursdays // 11:15- 12:00 and Fridays // 11:15- 12:05

Urban League of New England

Urban League

The goal of this project is to track and utilize voter turnout data to assist efforts in increasing voter turnout in upcoming elections based both at the local and national level, specifically in low propensity areas. First, the dashboard for Brokcton must be completed using historical and new data. Once this baseline is complete, the same dashboard can be created for Chelsea, Framingham, and Malden. The client is focused on communities of color and precincts with lower turnout. Our hope is that the social pressure information will show a discernible increase in those locations.

Use of Force in Waterbury Analysis

RACCE has received Use of Force reports from the Waterbury Police department that they would like to analyze. However, these reports were generated in image form (300+ reports in 3000 pages). The goal for this project is to generate a data analysis platform/tool that will allow individuals to use this data to empower social impact and desire to change public attitudes. A central part of this project is that it is important for people to understand not just the outcome of over policing but mechanisms that are in place that allow it to happen.

RAACE

Web Traffic Analysis

The organization aims to analyze web traffic data to understand the behavior, preferences, and engagement patterns of Gen Z and millennial users. With over 50 years of experience delivering counseling services, the organization seeks to modernize its service delivery and communication strategies to align with the needs of younger audiences. This project will focus on providing actionable insights into web traffic and user engagement from data such as Overall Traffic, Page Views, Active Users, New Users, Demographics, Geolocation,and Technology Devices so that can inform future decisions about improving the website's design, content, and functionality to better meet the needs of younger users and enhance service accessibility.

Credit.org

AIMpower

One of the focal points of AImpower.org is studying the role of technology in the marginalization of people who stutter. The organization has collected a dataset of stuttered speech recordings (72 people and a total of 50 hours) containing both conversational and voice command recitation speech, and have manually created verbatim transcriptions that include disfluent utterances. Furthermore, they’ve labeled the recordings on the different range of stuttering (mild to a lot), different kinds of stuttering (prolongation, filler words, long silence, repetition), and whether it’s freeform/casual conversation vs reading of voice commands. The goal of this project is to build off of Summer 2024, where they benchmarked these recordings against popular speech recognition models provided by OpenAI, Google, Facebook, Microsoft, and Otter.AI. You will quantify fluency bias in each of these speech models and visualize how the results change with different types of stuttering and across each model. *knowing chinese is not required, but knowing them may be beneficial

AIMpower.org

SWE · Fall 2024

ALS Resource App

The project aims to continue to develop a web and web app-based tool designed to assist patients in identifying their status and needs while also providing tailored recommendations and resources. Building on the foundation established by the previous team, the new development phase will focus on enhancing user experience by integrating feedback mechanisms, decision-making aids, and comprehensive content structures. The tool will prioritize the patient’s perspective, ensuring that the interface is intuitive and user-friendly.

BIPOC Voting

The project will focus on producing a data-driven investigation aimed at understanding the voting patterns of BIPOC voters in Massachusetts in the 2024 general election. The project will create a statewide map of precinct-level demographics (race + ethnicity and perhaps other attributes if time allows) to analyze precinct-level voting results from the 2024 general election.

WBUR

Broken Documentary

Millions of U.S. families lack health care, food and housing. And more than 400,000 kids annually are taken from their homes and put into foster care. But six in seven children are removed from their homes not for abuse, but for neglect, the blanket term that all too often is conflated with issues stemming directly from poverty such as the lack of food, shelter or childcare. Together we can fix the system - and save lives. The documentary represents the work of an unprecedented reporting team comprised of seasoned journalists with decades of experience covering child welfare nationally, intrepid student journalists doing enterprise reporting, and leading national experts working to identify and expose the underlying systemic problems in the child welfare system. By doing so the film will bring about real change to these broken systems. • https://brokenthefilm.org

State DCF

SWE · Fall 2024

BU Passport

Spark! and The BU Arts Initiative have been developing an app that tracks and promotes student engagement in arts events around campus. The app currently has 4 main features: • Events Page: The user can see upcoming art- related events. • Calendar: The user can view the entire calendar that uses an API to fetch data from CFA’s calendar • Leaderboard: Through a points system, the user can compare their engagement with the arts community with other users • Profile: The user can utilize their profile to track progress and events. BU Arts admin can also utilize this data as key metrics in assessing success of the application. By tracking utilizing either a photo upload or by GPS tracking, the Arts Initiative would provide prizes to winners determined by the backend. The app should also provide student demographic information (school, year, program, BU ID) for the annual BU Arts Initiative Report.

BU Arts Initiative

Contaminated Identification

This project, which started five years ago in Boston, combines an online platform, a data collection system, and a mobile app to analyze the content of bins based on visible elements. It is the team's longest-running project, beginning in 2021, and is led by a young startup of around 10 people. Spare-it has been a key partner of Boston University, where IoT devices and sensors deployed in the bins throughout the CDS building send real-time data every second to our system, which is then parsed and displayed on our admin dashboard for analysis. Additionally, students take weekly photographs of bin contents using the mobile application, and these images are manually labeled and added to the dataset daily. This combination of automated data collection and manual labeling allows for continuous improvements in waste analysis and reporting. Spare-it is now expanding the product's potential by integrating modern AI and machine learning techniques, focusing on neural networks for generating image segmentation models. During the Spring 2024 and summer terms, Spark! students developed a basic machine learning pipeline using the YOLOv8 framework by Ultralytics. Synthetic images were introduced to augment the dataset, though results were mixed. The model's performance varied significantly across 101 classes, with selected classes achieving around 68% accuracy overall. This semester, students will aim to further improve the model’s accuracy, building upon the insights from previous work. A key takeaway from earlier efforts is that enhancing data quality is crucial. The focus this semester will be on refining the accuracy of individual objects within the dataset, identifying the best labels, and fine-tuning the model to boost the precision of object detection and classification. Below are the key tasks as suggested by the client. Students do not need to complete all of them but should work with their team members, PM, Spark! staff, and instructor to come up with a plan and milestones for the semester. The goal of this project is to improve the existing contamination identification model, whether through dataset enhancement or model improvement.

Spare-It

DC City Government

Students will analyze patterns of 311 service requests in Downtown DC, focusing on patterns of request types, resolution times, and geographic trends. The neighborhood of interest is the Gallery Place Chinatown Area. The goal is to identify how service requests differ across neighborhoods, how they vary across neighborhoods, how quickly they are resolved., and how these trends have evolved over time.

Andree Entezari

SWE · Fall 2024

Editor AI

In this project, students will focus on integrating the AP Stylebook API into a text editing platform, aiming to streamline the process of applying AP Style guidelines in content creation. The primary goal is to build and optimize workflows that incorporate AI-driven style checking, ensuring that the tool accurately identifies and tags edits according to AP standards. Students will develop a prototype that enables users to connect their accounts and utilize the style checking features seamlessly. Additionally, they will research existing tools like Lingofy and Grammarly, evaluating their effectiveness and exploring the possibility of creating custom solutions. The project will also involve programming specific prompts into a Language Learning Model (LLM) and validating its accuracy with AP Style guidelines.

Michelle Johnson

Institute Diaspora Brazil (IDB)

Insituto Diaspora Brazil

Visualizing the Brazilian Diaspora provides interactive data analyzing the state of the Brazilian diaspora in the US to support and promote their initiative goals. The client is particularly interested in understanding how this data can reflect the economic and political influence of the Brazilian diaspora in the US.

Misconduct w/ Prosecutorial

• The goal is to automatically flag federal appellate decisions in which it seems likely the court found prosecutorial misconduct. • We want to build a database of all cases where misconduct was alleged, so we can start to see trends, such as prosecutors who repeatedly withhold evidence or areas in the country where there is more misconduct. • We don’t want the story to be about the numbers, though. Instead, we want you to report out the most egregious cases and any newsworthy trends you find.

The Intercept

Police Misconduct

• The goal is to automatically flag federal appellate decisions in which it seems likely the court found police misconduct. • We want to build a database of all cases where misconduct was alleged, so we can start to see trends, such as police who repeatedly withhold evidence or precincts/areas in the country where there is more misconduct. • We don’t want the story to be about the numbers, though. Instead, we want you to report out the most egregious cases and any newsworthy trends you find.

TBD

Data Science · Fall 2024

Quantum Cop

The police budget of approximately $400 million annually for the city of Boston is the largest expenditure for the city after education and represents approximately 10% of the city’s annual operating budget. Police overtime is guaranteed whether budgeted or not and regularly exceeds the budgeted amount. On average, 20% of the police budget goes to overtime. Many city officials would like to spend this money on more strategic investments and managing this significant unexpected annual expense, which has caused great uncertainty for the city. This project seeks to understand if there may be inherent inefficiencies in the system for managing overtime and the potential misuse of the overtime benefit by select police officers. There are three types of overtime: special events, detail, and court overtime. Many detailed rules in the police contract determine what officers are paid; for example, if an officer works for 15 minutes on a court overtime assignment, they are paid for 4 hours. Another rule states that they cannot collect overtime pay (1.5 times their hourly base rate) while working scheduled shifts or completing other overtime shifts. There are concerns that this may happen, and your project will investigate this possibility. You will aggregate overtime data collected from 2020 - 2022 for members of the Boston Police Department (data from 2023 may be added later in the semester) and merge these data with policing data represented by the Field Interrogation Observations (FIO) data set. FIOs are representations of every police interaction with a civilian. All of these records contain dates and times of day, which will allow you to identify potential overlapping shifts. For example, if an officer has a recorded FIO during a time when they were supposed to be working an overtime shift, this would be an indicator of noncompliance with department rules.

Yawu Miller

Racist Deeds

The Racist Deeds Project aims to expand the identification of property deeds with racist restrictions, initially focusing on Longmeadow, Massachusetts. These restrictions, targeting marginalized groups like African Americans, were outlawed by the Fair Housing Act of 1968 and locally in Massachusetts by 1946. This project seeks to streamline the identification process of discriminatory deeds to support the Longmeadow Historical Society. The focus of this semester is to build a data pipeline that will interact with deeds stored in a designated Google Drive location and potentially adjust some of the Gen AI prompts. The project will implement Optical Character Recognition (OCR) tools via a Python package to digitize and extract text from scanned deed documents, which will facilitate more efficient analysis and pattern recognition. Additionally, there will be an effort to refine and adjust the Gen AI prompts to improve the mapping (property). Data Pipeline Outline: 1. Data Sources Location: Google Drive Format: .tiff 2. Data Extraction Method: Download files from Google Drive Tools: Python, Google Drive API 3. Data Transformation OCR Processing: • Tool: OCR tool (e.g., Tesseract) • Objective: Digitize and extract text from .tiff files Content Analysis: • Tool: OpenAI (e.g., GenAI) • Objective: Identify if the document contains racist restrictions 4. Data Loading Output– if Racist Restriction Detected: • Extract information about the document’s location in the registry database • Extract the physical location of the property tied to the deed • Destination: Excel sheet or Google Sheet

MassMutual

Data Science · Fall 2024

ShotSpotter Political Discourse Analysis

• The goal of this project is to build on the Black Response’s campaign of ending ShotSpotter in Cambridge. Specifically, this semester we will focus on analyzing Cambridge City Council rhetoric surrounding the program while simultaneously continuing research questions (listed below) from previous semesters.

Shotspotter

Data Science · Fall 2024

STARS

This project aims to develop a system for monitoring and tracking sustainability-related research at Boston University to support the Sustainability Tracking, Assessment, and Rating System (STARS) Report. STARS is a transparent, self-reporting framework for colleges and universities to measure their sustainability performance, which BU reports to annually. The goal is to create a tool that consolidates data on research publications that address a Sustainability Challenge* , providing a clear picture of how engaged BU is in sustainability research across departments. The system will track faculty members' sustainability-focused publications by department. To ensure accuracy, the system will differentiate between entire papers focused on sustainability and those that only mention sustainability as a keyword, refining the focus to genuine sustainability efforts.

BU Sustainability

Wheelock / BPS Policy Help Bot

• High level description of the AI assistant’s capabilities (what will the AI assistant do? Internal vs. external) The project aims to improve the organization of public policy files by suggesting re-organization and developing a chatbot that can answer queries and direct users to the correct documents. • Use Case: Helping teachers access relevant data for parent-teacher meetings. • Admin: What are people searching for?

BU Hariri

ArtifexAI: Tech-focused Legislation Tracker / Predictor

ArtifexAI

Want to understand who really holds power in state politics and how laws actually get made? This project dives into the hidden networks of the Massachusetts State Legislature using real bill data from the past 5-10 years. Rather than just looking at official party lines, we’ll use network science, graphing algorithms and ML models to uncover the true alliances and power structures that shape our laws. By analyzing how legislators team up to cosponsor bills, we can reveal the unofficial ‘tribes’ within the legislature - who consistently work together, who bridge different groups, and who might be more influenced by special interests than party loyalty. Understanding the dynamics of legislative power and collaboration is critical for transparency and accountability. This project leverages Massachusetts State Legislature bill data (2017–present) to uncover hidden patterns in cosponsorships, voting behavior, and policy influence. By applying network science and graph-based methodologies, the team will identify key players, alliances, and power dynamics that may not align with traditional party lines. Using tools like graph analysis, clustering, and machine learning, the team will explore the following questions: • How do legislators align based on bill sponsorship or topics? • Are there influential legislators who shape bipartisan collaborations? • Can we identify outliers or groups influenced by funding sources or lobbyists? • Are there legislators who suddenly break from their usual allies on specific topics like climate change or healthcare? • Do certain groups only come together for particular types of bills? The deliverables aim to enhance transparency by producing actionable insights that advocacy groups, journalists, and citizens can use to understand legislative processes better.