The following completed and ongoing projects have been funded as part of the OSS Development Program (thanks to funds from the Alfred P. Sloan Foundation).
Completed Projects
User Interface platform that helps pull data from the NIH HEAL Data Platform and push it to the Qualitative Data Repository (QDR).
Project advised by Sebastian Karcher (QDR) and James Myers (Dataverse).
Personalized LLM Writing Assistant
This project focuses on training models to continue text in a scientific writing style and evaluating the quality of generated text through various linguistic and semantic measures. As part of this goal, the fellow developed a comprehensive pipeline for fine-tuning large language models (LLMs) and evaluating their performance using multiple similarity metrics and stylometric analysis.
Project advised by Johannes Himmelreich (Maxwell).
Tariff Document Digitization
An automated solution for digitizing complex tariff documents from the early/mid 20th century. Build as a Claude Skill that extracts structured data from scanned PDFs containing hierarchical commodity descriptions and associated tariff rates, transforming them into usable digital formats.
Project advised by Collin Capano and Will Gearty (SU OSPO) and Kristy Buzard (Maxwell School).
Preprint servers like the arXiv represent the bleeding edge of science, with up to hundreds of papers uploaded every single day. This natural-language based bot monitors a preprint server and notifies it’s owner if any relevant papers have been uploaded based on their research interests.
Project advised by Collin Capano and Will Gearty (SU OSPO).
Digitization and natural language analysis of the Montesinos tape transcripts from En la sala de la corrupción (In the Corruption Room). The recordings document meetings between Vladimiro Montesinos and various politicians and criminals during the Peruvian presidency of Alberto Fujimori.
Project advised by Jessie Trudeau (Maxwell School).
PERCEPT
A web-based speech perceptual rating platform developed at the Syracuse University Speech Production Lab. Researchers upload batches of audio recordings and trained raters listen and label them through the browser. Human ratings feed back into a machine learning classifier used to evaluate speech production accuracy over time.
Project advised by Jonathan Preston (Communication Sciences & Disorders) and Nina Benway (University of Maryland).
Ongoing Projects
galleryofpossibilities
A comprehensive QGIS plugin that provides a wide range of map icons with metadata display, designed for mapping applications. The plugin automatically downloads icons and metadata from Zenodo, ensuring users always have access to the latest icon library.
Project advised by Meghan Kelly (Maxwell) and Will Gearty (SU OSPO).
