GLPfit
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

GLPfit is a bilingual weekly-shot companion for Mounjaro and Wegovy users. Weekly injections create a weekly rhythm: appetite, nausea and energy change between shots. GLPfit organizes that routine around four daily records: weight, injections, meals and exercise. The demo follows Minji through shot day: log an injection with site rotation, see an estimated drug-level curve and remaining pen supply, record symptoms and weight, compare before/after meal photos to estimate protein actually eaten, and follow a gentle exercise routine. A Meta AI coach reads the shot schedule and server-computed trends, adapts its advice to the cycle, and remembers useful preferences. A printable monthly clinician report brings dose history, symptoms and nutrition estimates to the next clinic visit. Plate Lens combines SAM 3.1 food instance counts and bounding-box areas with Muse Spark vision and structured nutrition estimates. This turns segmentation into a practical measurement aid rather than a decorative overlay. All nutrition and drug-level values are estimates; the app does not diagnose, prescribe or recommend dose changes. Built as a React/TypeScript PWA with a FastAPI backend and Python agent. English and Korean are supported. The repository includes offline sample results, setup instructions and mocked automated tests. Research export is opt-in and de-identified.