Hi, I'm Kevin.

Working with Stanford AIMI on ML in health.

AIMI ↗

Circuit tracing medical vision-language models to find and suppress the reason they report findings they can't see.

Report ↗

$20k grant + $90k compute to build cool stuff.

Fellowship ↗

Arena-style benchmarking platform for clinical AI.

Platform ↗

A lightweight triage model for early detection of eye conditions. Acquired.

Archive

An explainable-AI (XAI) layer that shows and gives reasoning on which part of a scan a clinical model is actually looking at. Built on Grad-CAM and a fine-tuned MedGemma.

Archive

A desktop-native agent that reads your screen, predicts your next action, and executes in real-time. Open sourced.

Open Source ↗

Modular measurement device using optical rotary encoders. Best Presentation at Duke EGR101 Symposium; slated for $10M Duke accessibility project.

Research Poster ↗