heart rate74 bpm
modelconv1d-ae
statusnormal sinus
96% accuracy
1D conv autoencoder
WS live inference
- Trained a 1D convolutional autoencoder in PyTorch to learn healthy ECG morphology, so anything it reconstructs badly gets flagged as an anomaly.
- Built a FastAPI backend with JWT authentication and WebSocket streaming for live ECG inference, paired with a React 18 frontend.
- Integrated a Gemini-powered RAG chatbot that explains model predictions in plain language to non-technical users.
- Containerized the full stack with Docker Compose and automated build & deploy through GitHub Actions CI/CD.
PyTorchFastAPIWebSockets
JWTReact 18Gemini API
RAGDocker ComposeGitHub Actions