Take a trained scikit-learn model out of a notebook and serve it to real users as a reliable HTTP prediction service. You will persist models with joblib, wrap them in Flask and FastAPI endpoints, validate inputs, choose between batch and real-time inference, and version, monitor and containerise the result with basic MLOps practice.
Watch the free preview
Saving and loading models with joblib — free to watch, no account needed.
What you'll learn
- Serve a trained model to real users through an HTTP prediction service
- Persist and load models reliably with joblib and pin the environment for reproducibility
- Wrap a scikit-learn model in a Flask or FastAPI endpoint with proper input validation
- Choose between batch and real-time inference for a given workload
- Version, monitor and containerise a deployed model using basic MLOps practice
Syllabus
Persisting and Reproducing Models
Saving and loading models with joblibFree preview
Reproducibility and versioning your artefacts
Serving a Model over HTTP
Wrapping a model in a Flask endpoint
FastAPI with typed request and response models
Robust Prediction Services
Input validation and graceful error handling
Batch versus real-time inference
Production, Monitoring and MLOps
Versioning, monitoring and drift
An introduction to MLOps and reproducible deployment
Lab — 10 Exercises & Solutions
Exercises 1–5
Exercises 6–10