I'm a CS major at Cornell. This past summer I was at Meta, working on agentic ML workflows, benchmarking, and app start prediction models.

My work spans agents, ML research, and reinforcement learning. Recently I built Traceback, which red-teams RL agents to surface reward hacks that current solutions miss, and replicated Meta's Byte Latent Transformer, an entropy-based tokenizer-free LLM, from scratch at an ultra-low parameter scale.

At Cornell, I work at Hack4Impact, where I currently lead a team of 8 building the public impact dashboard for Medic's Community Health Toolkit, which supports 182K+ health workers across 24 countries.