4.8 million people in Bangladesh are affected by cataracts, yet fewer than 20% ever reach a specialist, rarely at an early, treatable stage. ALO is a clinical-grade AI app that detects cataracts from a simple smartphone selfie, requiring no specialist equipment or medical literacy.

The system uses a two-phase pipeline: a YOLO model isolates the eye region, followed by a DINOv2 vision transformer with CORAL ordinal regression to stage cataract severity. It achieves 90% training accuracy and 80-85% validation accuracy, and AUC scores of 0.91–0.94 across all severity classes.

Designed for home use, ALO enables early detection, automatic specialist referral, and has the potential to increase early diagnosis rates by 70%,  putting patients at the centre of AI-driven eye care and holding promise to reshape cataract screening worldwide.