NezT Labs: Private On-Device Receipt Scanner

NezT Labs (in-house concept build) · iOS & Android Apps · 2026

An in-house concept build: native expense capture apps for iOS and Android that read receipts entirely on the phone, so sensitive finance data is never uploaded unless the user chooses to export it.

What it does

A document camera with automatic cropping and perspective correction.

On-device text recognition reads the merchant, date, total, tax and currency, including BDT, GBP, USD, EUR, TRY and ZAR.

Rules plus a small on-device classifier suggest an expense category, and the user's corrections are remembered.

Export to CSV or PDF, or push approved expenses to accounting software through its API.

Engineering notes

iOS uses VisionKit's document camera and the Vision framework's text recognition. Android uses CameraX with ML Kit Text Recognition.

Built fully native on both platforms (SwiftUI and Jetpack Compose) for camera speed and smooth scrolling through long receipt lists.

Local data is encrypted at rest with the platform keychain and keystore, and the app works completely offline.

Tech stack

Swift, SwiftUI, VisionKit, Vision, Core Data, Kotlin, Jetpack Compose, CameraX, ML Kit, Room

Impact

Shows that finance capture can work fully offline, with nothing leaving the phone until the user decides to export.