EMMA Health
AI-assisted musculoskeletal health iOS app with on-device pose estimation, symptom triage and telemedicine.
EMMA — Efficient Musculoskeletal Management Assistant — turns a phone camera into a movement assessment tool. The patient performs a guided set of movements, the app estimates their pose in real time entirely on-device, and the resulting measurements feed a clinical assessment their provider can act on.
I worked across the ML and the app. On the model side that meant deploying MoveNet and YOLO through CoreML and TFLite, and tuning the inference pipeline until pose estimation held frame rate on mid-range devices without draining the battery. On the app side it meant the assessment flows, the virtual therapist module, video recording and playback, and a WebRTC telemedicine channel for live sessions with a clinician.
I also built an end-to-end NLP chatbot for symptom triage using spaCy and Hugging Face Transformers, handling intent classification and entity extraction against live user input.
Highlights
- Real-time on-device pose inference — no video leaves the phone during an assessment.
- WebRTC telemedicine so a therapist can review movement live.
- NLP triage chatbot for intent classification and symptom entity extraction.
- Shipped and maintained on both the App Store and Google Play.
Gallery
