Chinese Digit Classifier
A neural network built from scratch (no ML libraries) with a draw-to-predict web app.
- Python
- FastAPI
- React
- TypeScript
- NumPy

A data builder at the intersection of health & ML.
Data Science & Health Policy @ UNC–Chapel Hill
“Data is everywhere, but the glamorous part is the smallest part. You spend something like 95% of your time cleaning data and maybe 5% actually modeling it. I've also stopped believing good data science means hitting 90%+ accuracy. It means serving the people the data came from, and that starts with picking the right metric to measure in the first place.”
01 — Projects
Machine learning, healthcare analytics, and full-stack builds.
A neural network built from scratch (no ML libraries) with a draw-to-predict web app.
Reinforcement learning agents (DQN, DRQN, QR-DRQN, RND-DRQN) for a partially observable, windy grid world, with a full research paper.
Modeling the top drivers of diabetes from CDC survey data.
Predicting federal grant cuts across the UNC system with ML.
02 — Experience
Feb 2026 – Present
Boerwinkle Lab
Built a Python pipeline to classify resting-state fMRI components into brain networks, seizure-onset zones, and noise for pediatric epilepsy research.
May 2025 – May 2026
UNC School of Data Science & Society
Developed a LASSO-based risk screener from 650+ survey records; contributed to a CDC-funded study and was selected to present at NACCHO360 2026.
Jan – Sept 2025
UNC Water Institute
Built an interactive Tableau dashboard synthesizing 4,000+ studies on environmental health services for policymakers.
03 — Skills
04 — Beyond Code
Things I do in my free time: Gym, get outside, and rate meals on Beli. Currently hunting for good hiking trails around RTP (recs welcome).