OtoVision MLOps
End-to-end medical-imaging MLOps evidence and browser-side image QA workbench.
Author: Ankit Kumar Singh
Research portfolio demo: no autonomous diagnosis; no unsupported disease metrics.
Live image QA
Select a permitted, de-identified image. Processing stays in your browser; this Static Space does not upload it to an application server.
Choose an image to inspect dimensions, brightness and channel statistics.
End-to-end engineering path
Live now
Browser image QA
Portfolio documentation
CI-validated deployment bundle
Implemented in GitHub
FastAPI service
Docker packaging
Kubernetes manifests
Evaluation/provenance interfaces
Evidence boundary
Five-class checkpoint not configured
Disease metrics not claimed
Clinical validation not established
Architecture and source
The GitHub project retains the full training, service, container and orchestration implementation. A genuine release should pair an immutable checkpoint with saved predictions, split provenance and calibration/evaluation artifacts.