Introduction to Models as Web Endpoints

Predictive model as a web endpoint

In order for a machine learning model to be useful, you need a way of sharing the results with other services and applications within your organization. While you can precompute results and save them to a database using a batch pipeline approach, it’s often necessary to respond to requests in real-time with up-to-date information. One way of achieving this goal is by setting up a predictive model as a web endpoint that can be invoked from other services.

This chapter shows how to set up this functionality for both scikit-learn and Keras models, and introduces Python tools that can help scale up this functionality.