Model governance and compliance

The capabilities you will need for your ML pipeline to have some component MLOps life cycle processes:

• Version control
• Continuous integration/continuous deployment (CI/CD)
• Model management
• Scalable infrastructure orchestration
• Monitoring with alerting

These life cycle processes require some elaboration.

Version control is necessary to guarantee that a model can be reproduced exactly as it was intended. Often, notebooks are checked into git with data preparation steps, as well as an important set of recipes required to maintain the model over time. The model should always remain auditable. It also needs to maintain a change history and a record of all ML experiments run in its development. Docker images preserve much of the infrastructure needed to run the model. So, git and Docker are essential to track changes, collaborate effectively with peers, and reproduce all experiments.

Continuous Integration/Continuous Deployment...