Chapter 7: Feast Alternatives and ML Best Practices
In the last chapter, we discussed how to use Amazon Managed Workflows with Apache Airflow for orchestration and productionizing online and batch models with Feast. So far in this book, we have been discussing one feature store – Feast. However, there are a bunch of feature stores available on the market today. In this chapter, we will look at a few of them and discuss how they are different from Feast and the advantages or disadvantages of using them over Feast.
In this chapter, we will try out one other feature store, specifically Amazon SageMaker. We will take the same feature set that we generated while building the customer lifetime value (LTV) model and ingest it into SageMaker Feature Store and also run a couple of queries. The reason for choosing AWS over other feature stores such as Tecton, Hopworks, and H2O.ai is the easy access to the trial version. However, choosing the right feature store for you depends on...