Build the infrastructure and workflows that keep machine learning working after launch. This course covers orchestrated ML pipelines with Kubeflow, Airflow, and Prefect, experiment tracking with MLflow and Weights & Biases, model versioning and CI/CD, containerizing workloads with Docker and Kubernetes, and feature stores plus data versioning with DVC. You will finish with production monitoring for model performance and data drift, so your team can train, deploy, and iterate on models reliably at scale.

Modules

ML Pipelines and Orchestration

Lessons

• Why Notebooks Do Not Scale to Production Preview
• Orchestrating Workflows with Airflow and Prefect
• Kubernetes-Native Pipelines with Kubeflow

Experiment Tracking

Lessons

• The Case for Tracking Every Run
• Logging Runs with MLflow
• Collaboration Dashboards with Weights and Biases

Versioning Registries and CI/CD

Lessons

• Model Versioning and the Registry Pattern
• CI/CD Pipelines for ML Code and Models
• Deployment Strategies from Shadow to Canary

Containers and Kubernetes for ML

Lessons

• Docker Images for Training and Inference
• Running ML Workloads on Kubernetes
• GPUs Autoscaling and Resource Management

Feature Stores and Data Versioning

Lessons

• Why Features Need a Store
• Offline and Online Feature Serving
• Data Versioning with DVC

Production Monitoring

Lessons

• Monitoring ML Systems Beyond Uptime
• Detecting Data Drift in Production
• Alerting Retraining and Incident Response