Selected Projects
Below is a portfolio of my projects, collaborative work, and experiments. Each card captures the scope, timeline, and an at-a-glance summary. For more information, check out the Musings page for a deeper dive in to many of these projects.
-
Used graph analytics and artificial intelligence to generate explanations for predicting winners of college football matchups. Loosely inspired by My Team Is Better Than Your Team.
-
Leveraged the SPECTER embedding model to generate follow-up recommendations for a given research paper. Recommendations were drawn from a 60k+ paper dataset curated from arXiv.
-
Cacheus
Completed: May 2025Implemented the Cacheus cache replacement policy for the ZSim simulator. Experimented with various learning rate schedulers and judge combinations, yielding an approximate 5% improvement in miss rate over classical policies.
-
TechDisc
Completed: March 2025Reverse-engineered TechDisc's distance model for disc flight. Trained a GBM on a 20k throw dataset, achieving a sub-30ft error on the test dataset.
-
AggieAssign
Completed: Dec 2024Developed a full-stack application for generating class schedules for the computer science department at Texas A&M. Used linear programming techniques to design schedules maximizing instructor happiness and minimizing room waste. Reduced schedule creation time from multiple days to under a minute.