Summary

In this chapter, we focused on data drift and model drift in ML and how to monitor them using SageMaker Model Monitor and SageMaker Studio. We demonstrated how we set up a data quality monitor and a model quality monitor in SageMaker Studio to continuously monitor the behavior of a model and the characteristics of the incoming data, in a scenario where a regression model is deployed in a SageMaker endpoint and continuous inference traffic is hitting the endpoint. We introduced some random perturbation to the inference traffic and used SageMaker Model Monitor to detect unwanted behavior of the model and data. With this example, you can also deploy SageMaker Model Monitor to your use case and provide visibility and a guardrail to your models in production.

In the next chapter, we will be learning how to operationalize an ML project with SageMaker Projects, Pipelines, and the model registry. We will be talking about an important trend in ML right now, that is, continuous integration...