I build frontier AI systems — training large language models to reason and act.

I'm a member of technical staff at Amazon's SF AI Lab, where I work on reinforcement learning — CoT/RLVR training and agentic systems. Before this I sharpened computer-use agents at Adept, and over the past decade I've shipped ML at Stripe, Twitch, and Google — spanning risk and fraud detection, video delivery, and real-time ad bidding. Before all that I studied mathematics at UCLA and Northwestern.

My work lives at the boundary of research and engineering: I like taking open-ended questions, running the experiments to answer them, and turning the results into systems that ship.

You can read more in my CV.