Productionizing an online model pipeline
In the previous chapter, for the online model, we built REST endpoints to serve on-demand predictions for customer segmentation. Though the online model is hosted as a REST endpoint, it needs a supporting infrastructure for the following functions:
- To serve features in real time (we have Feast for that)
- To keep features up to date (we will use the feature-engineering notebook with Airflow orchestration for this)
In this chapter, we will continue from where we left off and use the feature-engineering notebook built in Chapter 4, Adding Feature Store to ML Models, in combination with a notebook to synchronize offline data to an online store in Feast.
The following figure shows the operationalization of the online model pipeline:
As you can see in Figure 6.14, we will use Airflow for the orchestration of feature engineering; data freshness...