Description
Summary We are HealthTruth (healthtruth.ai), a health AI startup in live beta: 600 active users, a 5,300+ waitlist, and 13.7M organic content views in 84 days with zero ad spend. Our founding team includes four leading physicians. Our AI, Neo, reads lab results against the ranges of optimal health and scores them into one number, the HealthTruth Score (300 to 1000). We are hiring our first engineer, and we are doing it the honest way: a paid one-week sprint first, then a path to a full-time founding role with equity if the fit is right on both sides. THE FIRST PROJECT: A ONE-WEEK PAID SPRINT Our core loop is lab report in, health score and charts out. The sprint is a hands-on review and hardening of that pipeline in our Supabase stack, with three deliverables: 1. Uploads work, every time. A lab report uploaded as PDF or photo is reliably extracted (AI-powered extraction) and captured into the database, with failures caught and surfaced, never silently dropped. 2. History is stored and charted. Labs from many different dates accumulate correctly per user and render as historical trends over time, not just a latest snapshot. 3. Bands are right. Every marker charts against our five bands (Optimal, Good, Fair, Improve, Focus) with correct ranges, colors, and placement, verified against our reference standard. One week, hourly, scope confirmed on a kickoff call. Deliver it well and the next sprints go straight at the launch roadmap: accounts, deeper personalization, public launch. This is the audition for a founding role. THE STACK TypeScript, React, Supabase (Postgres, edge functions), Vercel, and frontier AI model APIs. YOU - Have shipped production LLM features: prompt engineering, RAG, vector search, evals - Strong TypeScript and React, deep Supabase or equivalent Postgres - Treat sensitive personal data with the seriousness health data demands - Use AI tools daily to code (Claude, Cursor, or similar) - Want ownership, not tickets. Your trajectory here is tied to what you put in TO APPLY Send a short note with one production AI feature you built and shipped: what it did, how you measured its quality, and what broke. Links to live work or repos beat resumes. Full role description attached. The truth heals