Training Trains Internship
Student login
← MLOps learning path
MLOps · Module 8 of 10

Infrastructure for ML

Run ML workloads with containers, Kubernetes and cloud.

📖 Key ideas

  1. Docker packages models and dependencies.
  2. Kubernetes scales training and serving workloads.
  3. GPUs are scheduled as cluster resources.
  4. Managed platforms: SageMaker, Vertex AI, Azure ML, Databricks.
  5. Infrastructure as Code (Terraform) keeps setups repeatable.
💡 Real-life example: Containers are lunch boxes: the meal (model) arrives exactly the same wherever it is opened.
DockerContainer platform
KubernetesContainer orchestration
GPU nodeMachine with GPUs in a cluster
TerraformInfrastructure as Code tool