Training Trains Internship
Student login
← MLOps learning path
MLOps · Module 3 of 10

ML pipelines

Automate data preparation, training and evaluation.

📖 Key ideas

  1. A pipeline chains steps: ingest → validate → features → train → evaluate → register.
  2. Orchestrators: Kubeflow Pipelines, Airflow, Prefect, Vertex Pipelines.
  3. Each step is containerised and reusable.
  4. Pipelines can run on schedule or when data changes.
  5. Continuous Training (CT) retrains automatically.
💡 Real-life example: An ML pipeline is an assembly line where each station does one job and passes the result on.
Pipeline stepOne stage in the workflow
KubeflowML pipelines on Kubernetes
Continuous TrainingAutomated retraining
Model registryCatalogue of model versions