All posts

    Free 5-part course

    MLOps for Data Scientists

    From "it works in my notebook" to "it runs reliably in production," in five parts.

    The 5-part practical course on MLOps for data scientists: project structure, serving with FastAPI + Docker, experiment tracking with MLflow, CI/CD, and production monitoring. Each part is a guide with a working code repo.

    1. 1

      Notebooks Don't Ship: How to Structure Your ML Project So It Can Reach Production

      Why notebooks stall at the prototype line, and the scripts-first structure, locked dependencies, and isolated environment (uv) that make a project reproducible anywhere.

      Read Part 1
    2. 2

      How to Actually Ship Your Model: FastAPI + Docker for Data Scientists

      Wrap your model in a FastAPI service and package it with Docker so it runs the same on a server as on your laptop.

      Read Part 2
    3. 3

      Your Future Self Needs to Reproduce This: Experiment Tracking with MLflow

      Track every run with MLflow and use the model registry to promote or roll back versions without touching your service code.

      Read Part 3
    4. 4

      Deployment Isn't a Moment, It's a Pipeline: Automating Your ML Rollouts

      Build a CI/CD pipeline with GitHub Actions that tests, builds, and deploys automatically, even when a new model is promoted.

      Read Part 4
    5. 5

      How to Know If Your Deployed Model Is Actually Still Working

      The four layers of monitoring, data-vs-model drift detection with Evidently, and how to retrain or roll back when quality slips.

      Read Part 5