AI Lab / Open, runnable proof
Small scope. Serious execution. Every project exists to make one engineering capability undeniable.
Current experiments
Four compact systems, each its own public repo, three live on Vercel right now.
Multi-agent playground
Type a task — a Planner decomposes it, Researcher gathers context, Writer synthesises, Critic reviews. All four agents stream their thinking live through a real-time execution graph.
Model drift monitor
Trains a classifier, injects production drift, detects it with PSI and KS statistical tests, triggers automatic retraining, and versions every model in a JSON registry with a live dashboard.
LLM firewall
Paste any prompt — 64 layered heuristics screen it for injection attempts, PII leakage, jailbreaks, and unsafe content. Returns a 0-100 risk score, per-category breakdown, and a persisted audit log.
API architecture explorer
An interactive SVG node graph of a global identity and loyalty API. Click any node — gateway, OAuth, Redis, Postgres, queue — to read the design decision, trade-offs, and real config. Backed by a live rate-limited endpoint.
Project archive / Built over time
Selected public work from the path into senior engineering. Original interfaces, recast as a tactile project gallery.
Explore the foundations
View architecture ↗
