Summary

In this chapter, we looked at how models are deployed through SageMaker and covered how the endpoints are defined and invoked. Through the use of Spark's model serialization and deserialization, we illustrated how models can be shipped to other environments, such as a custom web service implementation in Flask. Finally, we outlined how your Spark model (or any other arbitrary model) can be served through SageMaker by registering a custom Docker image in AWS ECR.