I’m an independent researcher working on robot learning and evaluation, based in New York. I train vision-language-action policies on a low-cost SO-101 arm to study what they learn from different demonstration collection strategies, at fixed budgets, and I’ve also studied whether the standard transfer test for deception probes on language models can be trusted. Across both, I’m interested in whether the standard evaluations measure what they claim to. Before this I was a consultant at BCG, and I studied mechanical engineering at Cornell (B.S. & M.Eng.).