The problem
Small vintage boutiques sell mostly through Instagram messages. Listing each piece by hand (photos, size, condition, description) takes hours, and buyers can't tell how a coat will look on them.
What we built
- Lookli Studio for sellers: bulk photo upload, AI grouping of photos into items and draft listings with type, brand, size, condition, colour and material
- Background removal, cover images, short item videos and defect notes
- Shop pages with virtual try-on ("Essayer sur moi"), new-arrivals feed, reservations and seller contact
- Lookli Live: real-time auctions alongside a TikTok or Instagram live stream, with a 48-hour hold for the winner
- "Mon dressing": a private digital wardrobe with try-on and weather-aware outfit suggestions
- Staff accounts, admin review before a shop goes live, credits and shop-to-shop referrals
How it's built
One Cloudflare app serves the buyer site, the seller studio, the live auction rooms and a Telegram bot. Photos are resized on the seller's phone before upload, so uploads are about 20 times smaller, and AI work runs in the background so sellers can close the page.
Every risky feature sits behind an on/off switch, so it can be turned off instantly without a redeploy.
Numbers
- Launch rehearsal: 159 end-to-end simulated scenarios across shop, admin, account and credits, all passing
- AI listing cost below one cent per item, with zero invented values on the test set