Receipt Validation & Loyalty System
Alhamra Tower Receipt OCR
A receipt intelligence system that reads purchase receipts, verifies authenticity to prevent fraud, and awards loyalty points proportional to the verified spend — powered by an on-premise vision-language model.
01 — OCR
Vision-Language Receipt Reading
An on-premise Qwen 3.5 VL pipeline extracts merchant details, line items, totals, and dates from photographed receipts — handling varied layouts, lighting, and print quality without sending sensitive data to external APIs.
- On-premise VLM inference with Qwen 3.5 VL for data privacy
- Structured extraction of amount, merchant, date, and line items
- Robust parsing across different receipt formats and orientations
- End-to-end pipeline from image upload to validated transaction data
Vision-Language Receipt Reading
Core capability powering this project.
02 — Verification
Authenticity & Fraud Checks
Before any loyalty credit is issued, each receipt passes verification checks to confirm it is genuine — flagging duplicates, tampered images, and submissions that fail consistency rules.
- Duplicate receipt detection to prevent repeat submissions
- Visual and metadata checks for tampered or synthetic images
- Merchant and amount validation against expected patterns
- Reject fraudulent submissions before points are awarded
Authenticity & Fraud Checks
Core capability powering this project.
03 — Loyalty
Points Based on Verified Spend
Once a receipt is confirmed original, the system calculates and awards loyalty points proportional to the verified purchase amount — giving Alhamra Tower shoppers a seamless earn-and-redeem experience.
- Points scaled to verified receipt total after fraud checks pass
- Automated crediting without manual review for valid submissions
- Clear user feedback on approval, rejection, or points earned
- Designed for Alhamra Tower's loyalty and engagement program
Points Based on Verified Spend
Core capability powering this project.