Fig. 01 — Me, photobombing an Apollo Command Module.
01 / About
Half evaluator,
half roboticist,
fully caffeinated.
I build the systems that make frontier AI models better — from crafting evaluation datasets that probe real reasoning failures, to designing agentic benchmarks that stress-test long-horizon planning. Currently at Labelbox, I've helped grow one account from $0 to $25M ARR while pushing the state of RLHF and RLVR for the world's top AI labs.
On the robotics side, I spent years at Duality building closed-loop simulations, synthetic data pipelines, and digital twins using NeRFs and Cycle-GANs. I believe the best AI is shaped by people who understand both the math and the metal — the loss curves and the actuators.