London · Fontainebleau
GrapeSnap
Live · 2026 · a personal wine journal with AI label recognition
What it does
Point a phone at a wine label and get the bottle identified, its region and producer filled in, and space to add a tasting note. The core data - years of personal cellar records and tasting notes - sits in the vault as the source of truth and is never overwritten.
Why I built it
Wine apps either skew public (Vivino) or slow (notebooks). The interesting problem was building something fast enough to use at a tasting - snap, identify, log, move on - while keeping the personal archive private and structured enough to query later.
Where it stands
Architecture designed and data layer built; front-end development paused while NightDraft and the AI OS are the active builds. The bottle database is live and queryable.
Identifies wine from a label photo, checking on-device text recognition against a database before asking a vision model. Familiar labels take the direct route; the model handles the misses.
Stack SwiftUI, Apple Vision, Railway, Postgres, GPT-4o-mini