At Scale AI, I built Python tooling to author and validate agentic LLM benchmark tasks against a simulated company. I also wrote a deterministic verification gate that checks candidate SQL for schema conformance, data integrity, journal-entry arithmetic, and date constraints.
I ran benchmark tasks through a five-model review council and a six-trial pass@1 difficulty gate, delivering 81 QC-audited reference tasks across three business domains.
For SV Car Customs, I shipped a production e-commerce platform with Next.js, Hono, TypeScript, and Supabase PostgreSQL. I also built a reusable backend for multiple storefronts and integrated checkout, authentication, and purchase notifications.
