Skip to content


MLOps is a derivative of DevOps, specifically for machine learning. While the ideas and goals are similar, there are some crucial differences. Even if the code for a model does not change, additional data can mean a model needs to be updated. The lineage of a prediction is also quite different than with regular software development.

Machine learning lifecycle

Managing the full machine learning lifecycle is an important part of MLOps. Versioning, auditability, validation, deployment, and monitoring are all essential for a mature machine learning product.

By using Git as the single source of truth and a trigger to kickstart a number of validation steps, every derivative can be traced back to the version of the code and the pipelines that lead to a certain result.