AI Receipt Manager App & Web

I want to build a cross-platform system that lets any individual snap a picture of a receipt, have it read instantly by AI, then file, categorize, and turn it into clear spending reports—all from a phone or a browser.

Core flow
• On mobile (both iOS and Android) or the website, the user takes or uploads a receipt image.
• An on-device or server-side OCR layer extracts merchant, date, total, tax, and line items.
• The data is auto-categorized (groceries, fuel, dining, etc.) using a lightweight ML model, with manual override available.
• Users can view or export monthly and yearly summaries in PDF or CSV.

Tech preference
I am open to React Native or Flutter for the app and a modern web stack such as React + Node/Express or Django for the site. If you have a proven OCR/ML pipeline—Tesseract, Google Vision, AWS Textract, or your own TensorFlow-Lite model—mention it.

Deliverables
• Universal mobile app (iOS & Android) published to TestFlight / Google Play Internal Testing
• Responsive web dashboard deployed to a cloud host I can access (AWS, GCP, or Vercel)
• Source code in a private Git repo with setup docs
• Simple SQLite or cloud database schema with migration scripts
• PDF & CSV report generation module
• One-click backup/restore for all stored receipts

Acceptance criteria
1. Scanning a sample receipt auto-fills at least merchant, date, and total with 95 % accuracy.
2. Categories are pre-populated and editable; changes reflect in real-time reports.
3. Same user account stays in sync across devices within five seconds.
4. An end-to-end demo showing a receipt captured on mobile, reviewed on the web, and exported as a PDF.

Please outline your proposed tech stack, past work with OCR or expense apps, and an estimated timeline for an MVP.