ML platform
SageMaker Pipelines
Orchestrates repeatable ML workflows and CI/CD-style pipelines.
Key points
- Defines repeatable ML workflow steps such as processing, training, evaluation, registration, and deployment.
- Supports MLOps by making workflows versioned, trackable, and repeatable.
- Can integrate with model registry and approval steps for governed releases.
- Helps automate retraining and deployment workflows when data or code changes.
- Targets ML-specific pipelines rather than arbitrary application orchestration.
When to use it
- Choose Pipelines when ML workflows must be repeatable and automated.
- Use it for CI/CD-style training, evaluation, and deployment of SageMaker models.
- Use it when teams need lineage and approval around model releases.
Exam tips
- SageMaker Pipelines are ML-specific; Step Functions is general workflow orchestration.
- Pipelines do not label data; use Ground Truth for labeling workflows.
- Pipelines can include Clarify, Model Monitor, and registry steps as part of governance.
- Look for words like repeatable, automated, MLOps, and model approval.