to Present
LILAFounder
Solo-authored proactive AI assistant — a voice-first mobile product that reads signals from email, calendar, and workspace to decide when something is worth surfacing, and stays quiet otherwise.
- Sole author across the entire codebase — a pnpm/Turbo monorepo spanning a Node/TypeScript server, a Python/FastAPI AI service, a React Native mobile client and a Next.js site on Cloudflare Workers (~9,500 commits since February 2026; 9,550 at the 4 September 2026 count).
- Designed a 12-check delivery gate and a governed decision pipeline that scores whether a signal is worth interrupting the user for before deciding what to say.
- Built an evaluation harness that runs in CI and scores response quality, including a "should this stay silent" dimension via LLM-as-judge.
- Took the product to store-ready in sixteen languages — public site live in closed beta at hellolila.app, a 1.0 prepared in App Store Connect and a Google Play internal track, priced in 171 regions — and deliberately held the release until the proactive engine met its own bar.

