- There are tons of ML services that helps moving ml workloads to the cloud
- ML problems are quite different from the problems solution architects solve when working with the cloud
- ML related problems are typically more about the code & the models used, and not so much about the infrastructure (ex. containers)
- AWS dedicated ML exam
ML Services
- Self managed ML workloads
- Run own servers, install your own software on them, and run your own ml processes on top of these servers
- Can use servers(EC2 or EMR) or Containers (ex. ECS,EKS) and use these compute services to execute your ml code & build models
- can rent powerful instances (also GPU optimized) so you can run very complex ML computations in cloud
- good if you’re experiences
- SageMaker
- tooling for developing & testing ml models
AI services
- Helps with specific common problems you want to solve with AI
- ex. creating captions, analyzing images, etc
- uses a prebuilt ml model under the hood so you can use it without building ur own model