Implementing MLOps

DataRobot, through its MLOps suite, provides capabilities to enable users to not only deploy models in production, but govern, monitor, and manage the models in production. In previous chapters, we have looked at how models are deployed on the platform and using the Python API client. MLOps provides an automated model monitoring capability, which tracks the service health, accuracy, and data drift of models in production. The automated real-time monitoring of production models ensures that models have high-quality outputs. Also, when there is a performance degeneration, stakeholders are notified, so action can be taken.

In this section, we will focus on aspects of model monitoring that we didn't cover in Chapter 8, Model Scoring and Deployment, of this book. We looked at how to examine the quality of deployment services, as well as changes in the underline feature distribution between the training and prediction data across time through the service health...