Research
Bridging the gap between biological and artificial intelligence through advanced decoding and modeling techniques.
Conversational-Rate Brain-to-Text Decoding
Working at the Stanford Neural Prosthetics Translational Lab (NPTL), I develop systems that decode intended imagined speech from neural activity in real-time, aiming to restore communication for people with paralysis.
Real-time Decoding
Prototyped systems reaching 147 wpm in online demos, significantly faster than prior baselines. 11% WER offline.
Temporal Coding
Investigating how single neurons in orofacial motor cortex encode speech rate.
147 wpm streaming inference pipeline
Autoregressive modeling of organoid electrophysiology
Organoid Foundation Models
At Braingeneers (UCSC Genomics Institute), I build foundation models for cortical organoids to understand self-organizing neural computation.
SpikeData & Tooling
Co-leading SpikeData to simplify analysis of intrinsic neural recordings across large-scale HD-MEA datasets.
Closed-loop Stimulation
Applying models to RL-based goal-directed stimulation patterns in organoids.