Summary
This chapter introduced ML pipelines, illustrating their advantages in enabling consistency, correctness, and portability when implementing ML solutions.
An overview was given on the nascent MLOps field, a practice combining DevOps and ML to realize tested, scalable, secure, and observable production ML systems.
Further, we discussed the scikit-learn Pipeline class, a toolset to implement ML pipelines using the familiar scikit-learn API.
A practical, end-to-end example of implementing an ML pipeline for customer churn was also given. We showed how to create a scikit-learn pipeline that performs preprocessing, modeling, and tuning and is exportable for a software system. We then built a secure RESTful web API using FastAPI that provides an endpoint for getting predictions from our customer churn pipeline. Finally, we deployed our API to Google Cloud Platform using the Cloud Run service.
Although our deployment is secure and fully scalable, with observability, metrics...
