Kotlin Lottery Ticket Scanner

I need a production-ready Android application that scans printed lottery or gaming tickets—typically coming from an 80 mm kiosk printer—and instantly tells the user whether the monetary amount written in digits matches the amount spelled out in words. I’m building this under the name “Loterie Ticket Scanner,” and every line of code must live in my GitHub repository `loterie-ticket-scanner`.

Core flow
1. The user points any supported Android camera at the ticket.
2. Google ML Kit OCR (Text Recognition v2) captures both number and text fields in a single pass.
3. Kotlin code parses the two values, normalises currency symbols and separators, and performs an exact comparison.
4. If the values match, the screen shows a clear green confirmation; if they differ, the app immediately highlights the mismatch in red, drawing a bounding box around the offending field.
5. When characters are smudged, torn, or partially obscured, the app still attempts partial recognition and flags the uncertainty with a yellow warning overlay plus a short message explaining which part was ambiguous.

Technical notes
• Kotlin + Android Jetpack (CameraX preferred for compatibility)
• Google ML Kit on-device processing only—no network calls for OCR.
• Clean architecture (ViewModel, Use-case, Repository layers) so future features like barcode checking can drop in easily.
• Works from Android 8.0 (API 26) upward, tested on at least three form-factors (budget phone, Pixel-class phone, and tablet).
• Gradle build set up for CI; automatic unit tests for the parsing/comparison logic and at least one UI test using Espresso.

Deliverables
• Full Android Studio project pushed to GitHub `loterie-ticket-scanner` with an MIT licence header on each file.
• Signed release APK ready for Play Store internal testing.
• README covering build steps, device permissions, and a short troubleshooting section.
• Quick demo video or GIF illustrating a successful match, a highlighted mismatch, and the partial-scan warning.

Acceptance criteria
• 95 %+ recognition accuracy under normal kiosk-print quality.
• Highlight appears within 1 second of shutter.
• No crashes or major layout breaks on the test devices mentioned above.

Once merged to GitHub and the APK passes those checks, the project is complete.