Yosr — AI-Powered Accessibility Assistant for the Blind & Visually Impaired

Yosr is a fully voiced, Arabic-first mobile app (iOS & Android) that helps blind and visually impaired users independently handle everyday documents. Users can scan paper documents with AI-guided auto-capture, have forms analyzed and filled field-by-field through a voice-driven flow, check banknotes with an on-device currency reader, verify image quality before submission, generate formal letters and CVs, and listen to any document read aloud in Arabic or English.

Key highlights:

Accessibility-first engineering: custom native text and button overlays for flawless VoiceOver/TalkBack support, an Arabic phonetic layer so screen readers pronounce mixed Arabic/English content correctly, haptic and audio navigation feedback, and full RTL bilingual localization.

AI pipeline: on-device ONNX models for document segmentation and orientation detection, Arabic/English OCR, and a cloud backend (FastAPI + Celery) for form analysis, PDF processing, text-to-speech, and speech-to-text.

Production-grade stack: Firebase phone/Apple authentication with App Check, push notifications, offline local library (Hive), background task handling, and resilient camera/network error recovery.

Role: designed and built the full product — mobile app, AI integration, and backend API.