Building ML models and AI agents that solve real problems.

I’m most interested in AI that does more than chat: agents that go get the data, reason over it, and show their work.

At the Tech@NYU Shipyard Buildathon, a teammate and I built Cryptoshi, a crypto trading platform that blends Polymarket and Kalshi forecasts with Snowflake Cortex AI sentiment from X and a RAG chatbot that cites its sources. It won Best Use of Snowflake Cortex AI.

On my own I built receipts, an LLM pipeline that watches NFL betting markets, searches reporters’ posts on X when a line moves, and explains the move only from evidence it can cite. When the evidence isn’t there, it says so instead of guessing. I also built Foreseeer, a demand-forecasting site with a Gemini chatbot.

Right now I’m building SurroundSound with Tech@NYU, which syncs a room of phones into one surround sound system and uses ML to clean up each device’s calibration audio. The math side (statistics, probability, linear algebra) is what I’m studying at NYU Courant.