In one sentence
The three layers differ in ready-made capability, customization, low-level control, and operational responsibility.
Key points
- AWS AI Services: call pretrained APIs directly for the fastest path and least operations.
- SageMaker: use your own data to build, train, deploy, and monitor models.
- EC2/EKS/GPU plus frameworks: maximum control and maximum infrastructure work.
Exam focus
- Standardized and fast delivery points upward; customization and low-level control point downward.
Common pitfalls
- Do not choose by product name alone; confirm data type, latency, control, operations, and cost constraints.
Remember
Ready API → SageMaker platform → self-managed infrastructure.