Contributed to HealthMate with team Raatwaale Devs at Diversion 2K25. The prototype explores processing heartbeat audio and passing extracted features to a machine-learning model through a Python/FastAPI backend and Next.js interface. The submission describes Librosa audio preprocessing, frontend upload integration and model-inference work. It demonstrates an approach to audio-based health technology; it is not a clinically validated diagnostic tool.
Hackathon: Diversion 2K25 · Team: Raatwaale Devs
View the project on Devfolio · My Devfolio profile
Challenges
- Handling different audio formats and model input requirements; resolving frontend-to-backend upload errors.
Solutions
- Used audio feature extraction and standardized inputs, then refined multipart form-data handling and MIME types for the FastAPI upload workflow.