Computational neuroscience & ML, from organoids to conversational-rate speech BCIs
Research Tracks
My work focuses on two primary threads: decoding speech from the brain and modeling the dynamics of cortical organoids.
Speech Neuroprosthetics
Decoding Brain-to-Text at Stanford NPTL
Problem
Current speech BCIs are often too slow for natural conversation.
What I Built
A real-time streaming inference pipeline with low-latency text feedback loops.
Output
147 wpm imagined speech brain-to-text decoding reached in online demos at Stanford NPTL.
Organoid Foundation Models
Neural Data Science at Braingeneers/UCSC
Problem
Neural activity is complex and hard to predict, even in simple systems like organoids.
What I Built
An autoregressive foundation model that captures neural dynamics and predicts the activity of organoids.
Output
A single model that generalizes to unseen organoids and models dynamics; paired with an RL-based stimulation patterning algorithm to improve goal-directed training.
Professional Journey
ML Research Scientist Intern
Tacit AI / Reflex Technologies
Multimodal speech models & distributed training on B200s.
SURFIN Research Fellow
Stanford University (NPTL)
Real-time brain-to-text decoding & speech rate neural coding.
Undergraduate Research Engineer
Braingeneers, UC Santa Cruz
Foundation models for organoids & SpikeData toolkit.
Founder and CEO
TensorLearn
$1M+ funding & asset acquisition for AI education tools.
NASA Research Intern
NASA Ames Research Center
Transformer models for aviation service recommendations.
Neural Data Science Intern
UC San Diego (TNEL)
Songbird neural data analysis & NWB standardization.
Featured Projects
Explore some of my recent projects at the intersection of neuroscience and AI.
SpikeData
Python-based module for creating flexible and robust spike sorting pipelines for high-density neural recordings.
Polaris
On-device predictive maintenance for fishing vessels using sensor fusion and edge ML. 1st Place at Sushi Hackathon 2025.
StatScanner
iOS app that scans tables from paper and turns them into interactive stats visualizations using custom OCR + CoreML.
Sushi Agent
Autonomous LLM agents that discover and negotiate for private coupons. 2nd Place at Sushi Hackathon 2024.
Rubauto
Generative diffusion model for stylistic music synthesis and high-fidelity audio variations.