Scientific Projects
SkinX
An acne tracking app I built with a team of 4. You take one selfie and it finds and labels what is actually on your skin, then helps you keep track of whether it is getting better week to week.
- Fine-tuned a YOLO26 object detection model to locate and classify comedonal acne (blackheads and whiteheads) and inflammatory acne (papules, pustules, nodules), plus dark spots, from a single selfie.
- Built the Python data pipeline that merges multiple bounding-box-labeled dermatology datasets, so the model holds up across skin tones, genders, and photo quality instead of just one kind of face.
- Deployed the model on-device with CoreML and built the React Native/TypeScript app, with a Supabase backend for Postgres, auth, and row-level security.
- Wired up the Claude API to turn the detection results into a personalized care plan, barcode-based ingredient checks, and weekly progress tracking.
scoreexplorer.org
A web platform I built during my summer at GEn1E Lifesciences that aggregates and visualizes 129 clinical scoring systems used to track human health and disease. Researchers and clinicians can search and compare diagnostic scores across cardiology, neurology, oncology, infectious disease, and more. I designed the database and frontend so the tools are actually easy to browse and filter.