Amazon SageMaker 101

🔧 Fully-managed Machine Learning (ML) service
- 💡 Actually a collection of many features and products, bundled together by AWS
- Covers whole ML model lifecycle:
  - Fetch, clean and prepare data
  - Train, evaluate, deploy and monitor ML models
‼️ NOTE: in 2025, AWS rebranded SageMaker as "SageMaker AI”
- If you're looking for it in the AWS Console, look for it under that new name
- Just plain "SageMaker" takes you instead to "SageMaker Unified Studio" or the "SageMaker Platform," which is a wrapper around SageMaker and Amazon DataZone
- The exam only covers SageMaker AI at this time
Key features
- SageMaker Studio = IDE for ML model lifecycle
  - Build, train, debug and monitor ML models
- SageMaker Domain = isolation/grouping for a particular project
  - Contains EFS volume, users, apps, policies, VPCs…
- Containers = ML environments
  - Docker containers deployed to ML-tailored EC2 instance
  - Specific OS, libraries & tooling for ML workloads
- Hosting → Deploy endpoints for ML models
- ‼️ SageMaker itself has no cost!
  - …billed for created resources → complex pricing!!
  - 💡 ML resources are usually complex and high performant, it can get pricey! For extra reference, check: https://aws.amazon.com/sagemaker/pricing/
SageMaker Studio Screenshot