The best way to make the modules stick is to build something. These five projects take you from reproducible ML with Git and DVC through experiment tracking with MLflow, containerized model serving, CI/CD for ML, and finally monitoring a deployed model for drift.
Each one is a code-along video, so you can pause, type it out yourself and adapt it to your own ideas. Pick the project that matches the module you've just finished and come back for the rest.
Track datasets and trained models alongside your code with Git and DVC, so any experiment can be reproduced exactly.
Log parameters, metrics and artifacts for each run, compare results in the MLflow UI, and register your best model.
Wrap a trained model in a FastAPI prediction service and package it as a Docker image ready to deploy.
Build a CI/CD pipeline that tests your code and data, retrains the model, and ships the container on every change.
Build an Evidently monitoring dashboard that tracks data drift and model quality on a live model.