Mathematics, dance, and a decade of AI.

I am the founder and CTO of AutonomyAI, where we build agent harnesses that let teams ship production code into enterprise codebases. Before that I spent a decade building machine learning systems across audio, vision, text and vector search, and before that I was a mathematician who also danced.

I graduated from UC Berkeley at 18 with honors in theoretical mathematics and minors in computer science and dance, then got a master's in AI at UC San Diego. I built interactive projection systems for aerial dance, which turned into Zuzor, a company that put AI-powered computer vision into experiential media. We worked with NEC, Epson and Intel, and won awards I am still proud of. Then I led the development of neural networks for noise reduction at Waves Audio, including Clarity VX, which won an Emmy. I served as an elite academic officer in Unit 8200, leading AI research. Now I teach graduate AI and signal processing at Reichman University.

The through-line is the same thing I write about: the gap between the person who has the idea and the system that has to carry it. In dance it was choreography versus code. In audio it was a model versus a product. Now it is a PM, a designer or an engineer versus an agent that can build faster than the team can review.

I care about making that gap smaller, and about being honest when it does not work. Most of what I publish is a field report from a team that is still figuring it out, not a framework sold after the fact.

Outside work I train as an aerialist and acrobat. It is the only thing I do that cannot be rushed.