What's changed: Deepened MLS-C01 Chapter 4 (4 inference modes/MME/async limits/Neo, 3 drift kinds/Model Monitor 4 types/EventBridge retraining, 4 rollouts canary/linear/B-G/shadow/VPC mode+PrivateLink/KMS/Clarify + tables, scenarios, FAQ, traps; ja figures)
4.3MLOps and Security
Automation and safety—understand SageMaker Pipelines, Model Registry, safe rollout (A/B/canary), and security via IAM/KMS/VPC. Achieve reproducible, secure ML operations.
MLOps is about operating ML continuously and safely. Automate with pipelines, roll out safely, and protect with IAM/KMS/VPC.
4.3.1Automation, rollout, security
- SageMaker Pipelines: automate train→evaluate→deploy into a reproducible ML workflow.
- Model Registry: version and approve models and govern which is in production.
- Safe rollout: switch with limited impact via A/B (production variants)/canary/shadow.
- Security: protect data/models with IAM least privilege, KMS encryption, VPC/private connectivity.
Common on MLS-C01: automate train→deploy = SageMaker Pipelines, version/approve models = Model Registry, validate a new model on partial traffic = A/B via production variants (canary), and at-rest encryption = KMS, isolation = VPC, permissions = IAM least privilege. Keep training data private (no public S3), protected by VPC/encryption.
Continue reading — free sign-up
You're reading the free preview. Sign up free to read this section in full, plus every chapter (including 4+) and all questions.