Problem
Wearable telemetry hardware emits high-frequency cardiovascular and photoplethysmogram data streams (250Hz+). Traditional React Native architectures marshaling raw sensor arrays across the asynchronous JSON bridge suffer from severe frame drops, garbage collection stutter, and UI thread freezes during intensive rendering.
Role
Architect and primary mobile engineer. Designed the offline-first telemetry ingestion pipeline, custom native host objects, Skia rendering loop, and automated regression testing suite.
Architecture & Stack
- C++ JSI Host Objects: Bypassed the asynchronous serialization bridge to write sensor samples directly into shared memory buffers.
- Shopify Skia: Hardware-accelerated canvas rendering drawing smooth real-time ECG curves at locked 60fps.
- WatermelonDB & SQLite: Lazy-loaded local persistence queue syncing readings when network connectivity resumes.
- HealthKit / Google Health Connect: Dual platform synchronization maintaining background ingestion compliance.
Outcome & Metrics
- Maintained 60.0 fps rendering latency with zero UI thread hitches under sustained 250Hz sample streaming.
- p99 frame latency reduced from 42ms to 14.2ms.
- 99.98% crash-free session rate across 450,000+ active companion devices.