Offline AI billing and growth insights for India’s kirana stores, on the phone the shopkeeper already owns.
🏆 1st place, Paytm Build for India AI Hackathon (Track 1: Merchant Growth AI), out of 1000+ participants. Built solo as Team Velox; the first working prototype was built in 4 hours.
GinLo_demo_final.mp4
Download the Android APK (arm64, Android 8+) from Releases › v1.0-demo. It runs fully offline; no account needed.
- Learn: tap New product, name it (OCR can fill name and price from the label), capture 6 angles, Save.
- Bill: point the camera at the items, tap Add to bill, then show the UPI QR and mark it paid.
- Modify: type or say a change, e.g. “Maggi ke teen packet bache hain”; confirm with “haan”.
- Home: sales trends, top item, payment split and recent invoices (Products ⋮ › Add demo sales fills a month).
- Point-and-bill: the camera recognises and counts items against the shop’s own catalog; no barcode scanner.
- Learns any product: 6 photos per item, with augmented variants; OCR autofills name and price.
- Trends that grow sales: sales vs the last period, top sellers and stock left show what is worth buying.
- Easy ledger: every sale, UPI or cash, is logged with a shareable PDF invoice.
- Speak to modify: Hindi, Hinglish and English stock and price edits, always confirmed first.
- 100% offline, ₹0 extra cost: all AI runs on the phone; no cloud, no subscription, no extra hardware.
Flutter, Dart, SQLite, TensorFlow Lite (YOLOE detector, MobileNet-V3 embeddings), Google ML Kit OCR, on-device speech recognition, UPI QR (qr_flutter), pdf, fl_chart. Built with Claude Code.
See ARCHITECTURE.md and FLOWS.md for how it works.