— Digital Health
Clinical Vision Assessment Platform
Neurotrack
Neurotrack helps detect early neurodegenerative disease through validated cognitive and vision assessments taken on a phone. Over two years we architected and built the Android experience: a dynamic exam engine fed by backend APIs, camera-based pupil tracking during each step, and a fault-tolerant pipeline to upload ordered video segments—because a single error invalidates the entire clinical session and forces the patient to start over.
- 15+
- Exam types supported
- Strict order
- Video upload chain
- Camera-based
- Pupil tracking
The challenge
Each clinical exam combined timed stimuli, precise camera capture, and regulatory constraints. More than fifteen exam types each had strict sequencing and validation rules—a missed frame, out-of-order upload, or corrupted segment voided the whole assessment. Hard-coding every flow was unmaintainable, and patients could not afford to retake tests because of app failures.
Our approach
We designed a unified exam framework: the backend defines exam structure and UI via API responses, and the client renders flows dynamically while enforcing step order and preconditions locally. During capture, the app tracks pupils through the device camera, buffers segments with checksums, and uploads in strict sequence with retry and recovery logic. Failures surface immediately with clear guidance so users know whether they can continue or must restart—protecting data integrity for clinical review.
Results
Neurotrack could ship new exam types without parallel native releases, and patients completed multi-step vision assessments with confidence that recordings would survive real-world network conditions. The ordered upload pipeline kept invalid sessions rare and gave research teams trustworthy artifact chains for analysis.
— Start a project