KORTEX AI: scan-and-go checkout verification

2025AI / Computer vision

Hackathon project for checkout-anywhere stores: a phone scanning app, a camera vision service and a reconciliation engine that checks what shoppers scan against what the shelf camera sees.

KORTEX AI sign-in screen: scan-and-go checkout that checks itself
Computer VisionPythonFastAPIReactWebSocketHackathon
Live Preview Source Code

Problem

Scan-and-go shopping is fast, but stores lose stock to missed and mismatched scans. Most loss-prevention ideas treat the shopper as a suspect, which ruins the experience the format exists for.

Solution

Built with my team CIA TRIAD for the Everseen Computer Vision AI Hackathon. Shoppers scan real barcodes with their phone camera. A vision service (OpenCV, a YOLO detector and our own IoU tracker and shelf-zone logic) turns shelf camera footage into candidate pickup and return events. Our reconciliation engine matches scans to those events and decides whether to match, wait, ask or escalate, waiting a grace window for late scans instead of flagging instantly. A mismatch becomes a one-tap question for the shopper, and a retailer dashboard shows live sessions, discrepancies and an append-only audit trail. Checkout eligibility is enforced only on the backend.

Outcome

A working prototype deployed with a live retailer dashboard and customer app, eight demo scenarios for judges, and 96 backend and vision tests. It never uses facial recognition or stores video, and every event shows whether it came from a live camera, a simulation or the customer.