An interactive web-app that I wanted to make for my own chess progression. Interpretable Chess features a graph-based opening explorer using data from the Lichess Master's games, and tunable Stockfish engine analysis for open exploration. Individual positions and candidate opening moves give nuanced interpretations of opening motifs, tactical/positional considerations, and strategic goals, all generated by my own custom AI model.
An agentic system for early wildfire smoke detection. Four specialized LLM agents coordinate computer-vision tools (optical flow, SAM2/SAM3) and weather data to cut the false positives that fog and clouds cause in single-image CNN detectors. Evaluated on the FIgLib dataset. Graduate CV project (CSE 252D).
A browser arcade of single-player games. User accounts, a points economy, daily bonuses, unlockable cosmetics, and global leaderboards. Built with React, TypeScript, and Vite, backed by Supabase.
Final project for my Network Theory class. My partner and I analyzed the assist networks of all NBA players from 2001–2024, using data pulled from the NBA API. The visualizations can be filtered to specific teams or to all NBA players over a single year. Hovering a node surfaces player-specific statistics like assists, average teammates assisted, and centrality measures.
A Yelp-inspired social web app for university students, built during a 10-week San Diego Supercomputing Center summer 2024 internship. I collaborated with 5 other UCSD undergraduates using Agile, contributing across the front- and back-end. We shipped features for restaurant discovery, user reviews, and friend activity feeds tailored to a campus setting.