Built an AI grading/RAG platform with LLM routing; improved consistency and saved 60+ TA hours/term.
A winding path through ML, data, and engineering—each stop leveling up impact.
Built an AI grading/RAG platform with LLM routing; improved consistency and saved 60+ TA hours/term.
Shipped Java/Python microservices for NLP extraction across 20M+ docs; CI/CD on AWS for reliability.
Automated QA with Playwright + TypeScript; +45% regression coverage and faster release validation.



Realtime multi-LLM voice assistant (FastAPI, Docker); +35% command accuracy; 99.9% uptime.
GitHub RepositoryRealtime multi-LLM voice assistant (FastAPI, Docker); +35% command accuracy; 99.9% uptime.
GitHub Repository


Vector ‘Destination Genome’ + personalized ranking; +75% precision, 2.1× engagement.
GitHub RepositoryVector ‘Destination Genome’ + personalized ranking; +75% precision, 2.1× engagement.
GitHub Repository

Full-stack flower order management (React, Node, TS, MongoDB). AI recommendations lifted engagement ~25%; ops −40%.
GitHub RepositoryFull-stack flower order management (React, Node, TS, MongoDB). AI recommendations lifted engagement ~25%; ops −40%.
GitHub Repository


HMM (4-state) for OHLC trends (~93% accuracy); compared LSTM/ARIMA with ~20% efficiency gain.
GitHub RepositoryHMM (4-state) for OHLC trends (~93% accuracy); compared LSTM/ARIMA with ~20% efficiency gain.
GitHub Repository

NLP + word2vec that generates evocative promo ideas from features; ~30% marketing efficiency gain.
GitHub RepositoryNLP + word2vec that generates evocative promo ideas from features; ~30% marketing efficiency gain.
GitHub RepositoryInternship/job search platform (Flask + MySQL).
GitHub Repository