MLOps is a set of practices that combines machine learning, software engineering and operations to deploy, monitor and maintain machine learning models in production reliably and at scale.
In practiceMLOps extends DevOps to the model lifecycle, adding concerns such as data and feature versioning, model registry, automated retraining, drift monitoring and rollback. It is the operating layer that allows model risk management policies to be enforced consistently rather than depending on individual data scientists. A common confusion is to equate MLOps with a particular tool stack: the practice is defined by the controls it enforces, including reproducibility, traceability and the ability to retire a model cleanly.
A manufacturer running a predictive maintenance model across hundreds of machines uses an MLOps pipeline to retrain weekly on fresh telemetry, validate against a hold-out set, promote the new version through staging and roll back automatically if drift metrics breach thresholds.
This definition is maintained by Moweb partners and used in live client engagements. For how MLOps applies to your estate, or to challenge a working definition, speak to a partner.