All work
Computer Vision · Healthcare deployment · 2024
Patient Fall Detection on the Edge
Pose-estimation fall detection running in real time on NVIDIA Jetson devices inside hospital rooms.
01 The problem
Patient falls need to be detected instantly — without streaming sensitive video off-site.
02 What I did
- Trained and optimised pose-estimation models for fall detection.
- Deployed on NVIDIA Jetson edge devices with TensorRT for real-time inference.
- Kept all processing on-device to protect patient privacy.
03 Impact
- Real-time alerts at the bedside
- Privacy-preserving — video never leaves the room
- Low-cost edge hardware instead of GPU servers
Client work — details generalised to respect confidentiality. Happy to discuss specifics in an interview.