Selected data science, machine learning, and full-stack work — plus the experiments that keep the lab fun. Code lives on GitHub.
Full-stack NLP application for exploring events and entities extracted from the publicly released Epstein document corpus. A Python pipeline classified 2,599 events across 787 document threads and surfaced 515 unique entities; a React/TypeScript multi-view dashboard makes the corpus searchable.
Price model over 48,000+ NYC listings combining structured features with NLP signals mined from listing names (luxury keywords, capitalization ratio, word statistics). Benchmarked OLS, Ridge, LASSO, and Random Forest; the text features added real signal beyond location and room type.
A deterministic, trait-driven arena simulation: procedural biome terrain, weather systems, alliances and betrayals, and a narrative event feed. Started life as a Python prototype; now a TypeScript engine you can play right here — bring your own roster.
The lab itself: Next.js App Router, Prisma, and NextAuth with role-based admin, a markdown blog with live-preview editor, an arcade with per-user save states, and the photo-notebook design system you're looking at.