Chapter 12: Automating Machine Learning Workflows
In the previous chapter, you learned how to deploy machine learning models in different configurations, using both the SageMaker SDK and the boto3 SDK. We used their APIs in Jupyter notebooks, the preferred way to experiment and iterate quickly.
However, running notebooks for production tasks is not a good idea. Even if your code has been carefully tested, what about monitoring, logging, creating other AWS resources, handling errors, rolling back, and so on? Doing all of this right would require a lot of extra work and code, opening the possibility for more bugs. A more industrial approach is required.
In this chapter, you'll learn how to automate machine learning workflows with AWS services purposely built to bring repeatability, predictability, and robustness. Complex workflows can be triggered with a few simple APIs, saving you time, effort, and frustration. You'll see how you can preview infrastructure changes before...
