Skip to Main Content

Snaplog

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

Snaplog — Demo video

The problem A buyer meets dozens of vendors at a fair and comes home with a camera roll they can't search, notebooks of vendor and product details, and a spreadsheet to build by hand, matching every photo to the right note and the right item. The proof: a catalogue type is just a schema Snaplog extends the existing fabric catalogue with garments — and the garment category was added as a versioned schema definition, with no application code changes. Fabric is one schema (composition, weight, weave, colour). Garment is another (silhouette, construction, fabric used, sizing). The structure stays the same: Event → Vendor → Item → Variant. The model reads whatever the buyer captures against the schema that applies. extraction used to run on Gemini; this week we moved it behind a provider interface onto Muse Spark.

Team