Products we run in production. Real users, real stakes.
Five Claude agents working one incident together: triage, logs, code, hypothesis, routing. It took 3rd place at my first hackathon and helped me land the R&D role I have today. Everything else in this alley traces back to what I learned building it.
Clinical practice manager for private-practice psychologists in México. Patients, sessions, tasks and an AI assistant — built with a real pilot therapist.
AI governance proof-gap auditor. Scans your AI system against NIST AI RMF, ISO 42001 and the EU AI Act, and shows exactly what evidence you're missing.
Clean-room reverse-engineering skill for Claude Code. Understand any repo's architecture with a knowledge graph, interrogate the design, then rebuild it your way. Don't fork the code — fork the design.
Pre-deploy security gate for vibe-coded Supabase apps. It reads your migrations and RLS policies locally and blocks the deploy on the leaks that matter — runs as a CLI, a Claude Code skill, and a GitHub Action.
A clonable personal AI operating system template — the same pattern behind our own daily operations, packaged to teach.
Local, private dictation polish. Raw speech-to-text in, clean ready-to-paste text out — bilingual Spanish/English, cleaned by a small model that runs entirely on your machine. Zero dependencies, and your voice never leaves the device.
