Scikit-Learn Inference API Development

I have a set of Scikit-learn classification models ready and now need a lightweight, production-ready API that can expose real-time inference endpoints. The core purpose is simple: receive JSON input, feed it to the appropriate model, and return the prediction as quickly and reliably as possible.

You’re free to choose the server stack (FastAPI, Flask, Django REST, or something comparable) as long as it is Python-based and easy to containerize. Clean code, clear separation of concerns, and solid error handling matter more to me than framework loyalty.

Key deliverables I expect:

• A RESTful (or GraphQL if you can justify it) service that loads one or more pre-trained Scikit-learn classification models and serves /predict and /health endpoints
• Simple input validation plus confidence/probability output when available
• Dockerfile and brief README covering setup, environment variables, and run commands
• Unit tests demonstrating that the API returns the correct class labels on a provided sample set
• Minimal but clear inline documentation so future models can be swapped in without rewriting core logic

Acceptance criteria: the container must start, hit /health and return 200 OK, then accept a POST to /predict with my sample payload and echo back the expected class within 200 ms locally.

If this sounds straightforward to you and you’ve shipped similar inference services before, let’s get this running.