I build AI systems that turn frontier research into business value.

I’m a Founding Engineer building AI systems that can understand the real world over time.

At FlowState AI, I build video intelligence systems for businesses. Underneath that work is a research problem I’m deeply interested in: long-form and cross-video understanding, where models need to follow context, connect events, and reason over hours of footage or many related videos.

Much of my work has lived in settings where intelligence has to survive contact with messy reality: reconstructing 3D scenes from sparse imagery at MIT Lincoln Laboratory, building multimodal personality detection at Openstream.ai, and developing GraphSLAM for Carnegie Mellon’s autonomous racecar as part of the team that completed the first fully autonomous Formula-style lap in North America.

I’ve also spent years competing in chess, reaching a national ranking of #48 in the U.S. among 18-year-olds. My TEDx talk came from a different side of that world: lessons learned from playing street chess hustlers around the world. That experience shaped how I think about pattern recognition, adaptation, and decision-making under pressure.

Across these experiences, I keep coming back to the same idea: intelligence is most interesting when it has to operate in dynamic, imperfect environments. I’m working toward harder AI research problems by building production systems from the ground up: the data, infrastructure, evaluation, and product loops that make models useful outside the lab.

Long term, I want to help build AI systems that understand the real world over time, across video, multimodal data, and other messy streams of experience.