I want to launch a web-based bill-scanning application that uses AI/OCR to read purchase and sales bills. Visitors will be able to upload a photo or PDF, have the system automatically extract all key data (supplier, date, items, totals, taxes), categorise each expense on the spot, and finally export the cleaned-up result to a nicely formatted PDF they can download or share. Core features I need built in: • Automatic data extraction driven by a reliable OCR or vision API • Real-time expense categorisation with editable fields • One-click export of the parsed bill to PDF Everything has to run in the browser—no native iOS or Android builds for now—so a responsive interface and a lightweight back-end API are essential. Please structure the code so I can eventually extend it into a full SaaS (user accounts, billing, etc.). Deliverables will include: 1. Fully functional web app deployed on a test server 2. Source code with clear documentation and setup scripts 3. Brief report on the AI/OCR model or third-party service you selected and why 4. Basic unit/integration tests proving that extraction, categorisation and PDF export work on a representative sample of purchase and sales bills If you’ve already worked with tools like Tesseract, Google Vision, AWS Textract or similar, let me know; a pre-trained, production-ready approach is preferred over starting from scratch. I’m ready to get this moving quickly—tell me how you’d tackle the data extraction accuracy and the front-end workflow, and include any relevant demos or repos you can share.