Neuroscience-trained physicist bringing brain-decoding methods to interpretability.
Why chosen
He connects reading brains with reading neural networks. During his Stanford Applied Physics PhD he built EXTRACT, a robust method for pulling individual neurons out of calcium-imaging data, and co-authored CORNN, a NeurIPS 2023 method for fitting recurrent networks to neural recordings at speed. In November 2026 he joined Goodfire, the interpretability company. From neurons to features.