Jacy Weems, Emmanuelle Belanger

Health Services, Policy and Practice, School of Public Health

There are several approaches to qualitative content analysis, a widely used and flexible research method for describing textual data. Qualitative content analysis may yield rich contextual evidence around the phenomena of study, but it is often limited by the manual effort required to analyze large-scale data. The “Qualitative Content Analysis Using R and Large Language Models” project establishes a new educational pipeline that integrates Large Language Models (LLMs) with R-based workflows, enabling health and social science researchers to use generative AI as an aid to content analysis of large amounts of data in the programming environment they use most.

Grounded in the analysis of high-volume public datasets - in this case, nursing home inspection records - the project focuses on moving beyond simple automation. The project will develop workshops and open-source materials that teach scholars how to select models, evaluate assumptions, and interpret outputs with a focus on reflexivity and rigor. By maintaining a “human-in-the-loop” approach, this initiative empowers the interdisciplinary Brown community to analyze extensive text sources more efficiently while preserving the interpretive depth of qualitative content analysis through manual review and input.

Laura Smithson, Chris de Graffenried, Ana Sofia Cepeda-Diaz

BioMed / Molecular Microbiology and Immunology

The “Creating the Trypanosoma brucei digital cell” project targets Trypanosoma brucei, a parasite responsible for severe illness in both people and livestock across sub-Saharan Africa. The parasite’s ability to move and infect its host depends on its unique shape, which is held together by a rigid “skeleton” of polymers called the subpellicular microtubule array. Because these microtubules are packed just 18-22 nanometers apart, it has been historically impossible to map their full organization.

Using a high-resolution 660-gigabyte imaging dataset and a specialized deep learning model, the team is building the first 3D “digital cell” to track how this microtubule array is remodeled and partitioned during cell division. By providing a high-resolution map of this parasite-specific structure, the project aims to identify essential proteins that could serve as targets for new medical treatments, while sharing the resulting data as a massive, open-access resource for the global scientific community.

Margot Jackson, Jiwan Lee, Michael Silverstein and Lauren Schlichting 

Sociology

While international research suggests that paid family leave significantly improves family well-being, rigorous causal evidence from U.S. state programs remains scarce. The “Leveraging Rhode Island Linked Administrative Data to Understand the Effects of Paid Leave on Children’s Health and Educational Outcomes” project utilizes the Rhode Island Child Health and Education Study (RICHES) – a massive administrative dataset linking records for over 340,000 children – to investigate how the state's 2014 Temporary Caregiver Insurance (TCI) program has impacted developmental trajectories.

Using a Regression Discontinuity in Time (RDiT) design, the research team will compare outcomes for children born just before and after the implementation of expanded leave. By tracking indicators ranging from immunization rates and neonatal health to third-grade math and reading proficiency, this study aims to provide a definitive account of how access to extended paid leave reduces educational inequities and supports early childhood development. This work bridges the gap between data science and public policy, providing actionable insights for the future of family-centered legislation.

Abebe Fola

BioMed / Pathology & Laboratory Medicine

Malaria remains one of the world’s deadliest diseases, and emerging resistance to primary treatments (ACTs) in East Africa threatens to reverse decades of progress. The “Developing an Oxford Nanopore Whole-Genome Sequencing and Analysis Pipeline for Plasmodium falciparum Drug Resistance Surveillance” project addresses a major technical hurdle: the malaria parasite’s genome is notoriously difficult to “read” because it is highly repetitive and complex. To overcome this, the project is developing a specialized analysis pipeline using Oxford Nanopore long-read sequencing, which allows for a much clearer and more complete view of the parasite's genetic code than traditional methods.

By applying this new pipeline to patient samples from Ethiopia, the project will search for novel “structural variants” – large-scale genetic changes – associated with drug resistance that current surveillance tools miss. This project not only aims to identify new markers for drug tolerance but also focuses on open science and global capacity building. By sharing the open-source code and training collaborators in Africa, this initiative creates a scalable, real-time surveillance system to track and contain resistant malaria lineages before they can spread.

Lindsey Kaler, Cameron Arnzen, Niamh Stull and Susan Moffit

Watson School for International and Public Affairs

The project “How Does the Dissolution of the US Department of Education Impact the Education Ecosystem” investigates the structural consequences of the March 2025 dissolution of the US Department of Education, treating the event as a pivotal case study for the future of American “bureaucratic capacity.” To understand where federal expertise goes when a department is eliminated, the research team is leveraging a massive, novel dataset of universal LinkedIn employment records. By tracing the trajectories of displaced employees, the study will map how specialized knowledge is redistributed across state agencies, local school districts and private organizations.

Using advanced statistical methods, including event-study and “difference-in-differences” models, the project aims to quantify the impact of this policy shift on the broader education labor market. This research not only addresses the immediate future of the US education ecosystem but also serves as a critical pilot for understanding how large-scale reductions in the federal workforce might impact other vital sectors, such as public health, environmental regulation and scientific research.

Michael Satlow

Judaic Studies

The “The Jewish Studies Citation Network” project addresses a fundamental challenge in the digital humanities: the extraction of structured bibliometric data from the highly unstructured and diverse citation styles of humanistic journals. By utilizing local Large Language Models (LLMs) to process over 35,000 full-text articles from JSTOR, the research team is constructing a massive citation network that maps the intellectual architecture of Jewish Studies.

The team will refine this data by accounting for complex variables, such as Hebrew-language references and primary source citations, before converting the network into an RDF format for analysis in Neo4j. Beyond mapping a single field, this project develops a generalizable pipeline and a set of open-source tools that can be adapted by other humanistic disciplines to mine semi-structured texts, ultimately transforming how we visualize and understand the global movement of ideas.

Jenny Zhu, Savvas Koushiappas, Ayaan Jamal, Bruno Rodriguez-Mendez, Ryan Pang

Applied Math / Computer Science / Engineering

The “Bloom: A Smart Plant System Integrating Sensor-Driven Data Analytics to Promote Urban Sustainability and Human-Plant Interaction” project is an innovative Internet-of-Things (IoT) platform that utilizes a multi-sensor array-measuring light, soil moisture, air humidity and temperature to optimize plant health through real-time data analytics. At the heart of the system is an adaptive database that dynamically adjusts its predictive models based on the specific requirements of different plant species, providing users with personalized, data-driven care recommendations via a web and mobile dashboard.

This project moves beyond simple automation to explore sustainable human-plant interaction. Following the successful development of a functional prototype, the team is now scaling the system to include solar-powered, self-sustaining hardware. By open-sourcing the code and publishing anonymized environmental datasets, Bloom will serve as a valuable resource for research and education in environmental modeling, data science and human-centered design.

Ece Uzun, Jeremy Warner, Seema Nagpal and Samuel Rubinstein 

BioMed / Pathology & Laboratory Medicine

The “iPREDiCT: A toolkit for cancer recurrence prediction using integrated genomics and clinical data” project is a multi-modal deep learning framework designed to enhance the accuracy of cancer recurrence forecasting by integrating clinical variables with complex genomic data, including DNA-sequencing, RNA-expression and methylation profiles. While many existing models suffer from small sample sizes or a reliance on limited data types, iPREDiCT utilizes attention-based neural networks to synthesize diverse biological and geo-spatial factors. Initially focusing on glioma, multiple myeloma and breast cancer, the project will culminate in a user-friendly, stand-alone executable application. This architecture allows for real-time, HIPAA-compliant predictions on local machines, bridging the gap between sophisticated computational biology and point-of-care clinical decision-making.

Rebecca Hubbard, Tiffany Hsieh

Biostatistics, School of Public Health

The “Data Science Collaborative for Real-World Evidence in Cancer Therapeutics” project brings together experts from medicine, public health and computer science to better understand how different chemotherapy treatments perform in the real world, particularly for older breast cancer patients who are often underrepresented in traditional clinical trials. By analyzing cancer registry data linked to national Medicare claims, the team will investigate whether a “de-escalated” treatment plan can be just as effective as more intensive regimens while reducing the risk of heart-related side effects.

To achieve this, the researchers are developing advanced AI models capable of “reading” patient histories to identify characteristics of patient health status that aren't standardly available in medical records. These insights will be used to create an “emulated clinical trial,” providing decision-makers with high-quality, data-driven evidence to help them choose the safest and most effective treatments.

Viktor Gavrielov, Amanda Khoo

Center for Career Exploration

The “From Engagement to Outcomes: Predicting and Improving Student Career Readiness at Brown” project creates a sophisticated data-science bridge between current student behaviors and long-term career outcomes. By integrating “rich engagement signals” from platforms like Handshake and BrownConnect+ with results from Brown’s First Destination Survey, the team will develop a “Career Readiness Index” (CRI). This index uses normalized metrics and causal analysis to move beyond correlation, identifying which specific sequences of activity – such as early advising, career fairs or specific workshop types – actually drive career readiness. The resulting analytics pipeline and no-code dashboard will allow the Center for Career Exploration to perform fairness audits and implement proactive, personalized interventions for unengaged student populations, fostering a more equitable transition from campus to career.

Richard Gaitskell, Gaetano Barone, Ian Dell’Antonio, Matt LeBlanc, Jonathan Pober and Jennifer Roloff

Physics

Scientific discovery is reaching a computational inflection point where the rate of data generation is outstripping our ability to store and process it. The “Brown Efficient AI Compression Network (BEACON)” project is a collaborative initiative designed to bridge the gap between domain-specific physics and advanced machine learning. By funding a cohort of interdisciplinary student researchers, BEACON will develop cutting-edge compression algorithms – including variational autoencoders and graph-based neural networks – to represent complex data more efficiently.

The project targets five distinct research frontiers: LHC particle point clouds, dark matter signal traces, astronomical telescope images, radio-interferometry spectra and high-granularity silicon device waveforms. Each of these modalities requires a “physics-informed” approach to lossy compression, aiming to reduce data volumes by 1 to 3 orders of magnitude while preserving high-dimensional correlations. By fostering cross-pollination between Physics, Computer Science and Applied Math, BEACON will create a unified toolkit for the next generation of data-intensive scientific exploration.

Kenya Andrews

Biostatistics, School of Public Health

The development of “The Machine Learning Conference on Justice in Education, Healthcare, and Public Policy” is an initiative dedicated to redefining how algorithms are designed and deployed in life-critical sectors. Moving beyond traditional “fairness metrics,” which often rely on unrealistic data assumptions, this conference focuses on a justice-oriented framework that prioritizes removing systemic barriers. By convening a multidisciplinary group of researchers, clinicians, and policy experts, the event will feature peer-reviewed presentations and a 24-hour student hackathon. Participants will use real-world datasets to develop ML solutions that are not only technically accurate but are rooted in the specific needs and contexts of the communities they serve.