Your complete guide — from development environment setup to a production-grade MLOps platform. All projects run locally inside VirtualBox (Ubuntu 24.04 LTS).
Course Learning Path
Follow projects in order — each one builds directly on the previous. Never skip a Setup lab; they install the tools you need for the next chapter's practice projects.
🔧 Setup Projects
Install Python, Git, Docker, and MLflow inside your Ubuntu 24.04 VirtualBox VM. Verify every tool before the course begins.
→ Setup LabInstall DVC, connect it to a local remote folder, and confirm push/pull works before the data-versioning chapter.
→ Setup LabRun Prometheus and Grafana via Docker Compose and install Evidently AI before the observability chapter.
→ Setup LabInstall Minikube and kubectl inside the VM to run a local Kubernetes cluster — no cloud account needed initially.
→📚 Practice Projects
Analyse a real ML production failure, map it to the ML lifecycle, and recommend MLOps tools that would have prevented it.
→ Practice ProjectCreate a professional ML repository, simulate a full branching workflow, resolve a merge conflict, and document a versioning convention.
→ Practice ProjectVersion two dataset iterations with DVC, run 3 MLflow experiments, compare them in the UI, and promote the best model to the Registry.
→ Practice ProjectRefactor notebook code into modular pipeline components, wire them into a Prefect DAG, and run Optuna hyperparameter tuning.
→ Practice ProjectBuild a FastAPI prediction endpoint, validate requests with Pydantic, and containerize the full application with Docker.
→ Practice ProjectWrite a GitHub Actions workflow that auto-tests, builds, and pushes the Docker image to a registry on every push to main.
→ Practice ProjectLog predictions to Prometheus, build a Grafana dashboard, run Evidently AI drift reports, and configure an alert rule.
→ Practice ProjectWrite a model card, test for bias across a subgroup, apply SHAP to explain predictions, and document a privacy/access-control plan.
→ Practice ProjectDeploy the containerized API to Minikube, configure auto-scaling, and perform a zero-downtime rolling update.
→ Practice ProjectCompare MLflow, Kubeflow, and Metaflow against defined criteria and recommend a stack for a hypothetical 15-person team.
→🏆 Final Capstone