An approach to the management of the lifecycle of machine learning model development, model operationalization, and model iteration and versioning. As an approach, ML Ops provides both methods as well as tools to make sure that ML models continue to provide results at required levels of performance. There are two major components of ML Ops:
DevOps for ML
- Model development lifecycle
- Model deployment
- Model versioning
IT Ops considerations for ML
- Model-specific Ops
- Measuring and managing Model drift
- Measuring and managing Data drift
- Establishing data provenance
- Establishing Data and Model Governance
- Enabling Model discovery
There are becoming too many “Ops” with vendors making distinctions between ML Ops, Model Ops, AIOps (which has nothing to do with ML), xOps, and more. Ignore the terminology battles. Listen to our podcasts on this topic.
