In short
MLOps is the set of practices that makes developing, deploying, monitoring and updating machine learning models repeatable and auditable.
A model that performs well in the lab will not necessarily keep performing well in the field. As data distributions shift, model performance can degrade.
MLOps covers data and model versioning, automated training pipelines, deployment, performance monitoring and retraining when needed. In industrial settings these processes must work both in the cloud and on edge devices.