Alex Gomez

A Next.js + TypeScript web app deployed on Vercel, with a sibling React Native / Expo mobile client sharing lib, prompt, and type layers on top of a Supabase Postgres data tier with versioned SQL migrations. Identity-aware data access is enforced through Row Level Security (auth.uid() = user_id) with split service-role vs. authenticated grants. Server-only secrets are kept out of mobile and browser bundles via strict EXPO_PUBLIC_* separation. A per-IP token-bucket rate limiter (8/min) shields Gemini-backed API routes, with Sentry + PostHog observability and a model-fallback path for 429s. Tests run on a mocked Gemini SDK.

Built and benchmarked three NLP classifiers — TF-IDF + logistic regression, a 1D CNN, and a BiLSTM — for IT support-ticket routing over ~11,920 English tickets. Owned the full pipeline from data cleaning through model evaluation and documented results in an IEEE-format paper. Same supervised text-classification pattern that powers automated alert and event triage in production security tooling.

Recommendation system that helps users discover manga through an interactive quiz and swipe-based interface. Built a TF-IDF + cosine similarity model from scratch in scikit-learn for personalized recommendations, and integrated the Google Gemini API so users can chat with AI personas of their favorite characters. Full stack: React frontend, Flask REST API, and SQLite for user preferences and inventory. Submitted to ShellHacks 2025.

Designed a normalized PostgreSQL database and ETL pipeline in Pandas to ingest 30,000+ real-world charging sessions from Caltech. Tuned geospatial queries with PostGIS and B-tree indexes, cutting execution time ~60% for complex aggregations.