Glossary
MLOps (model operations)
Building, testing, deploying and watching machine learning models with the discipline of software delivery: pipelines, versioning, registries, monitoring.
The MLOps pipeline is where models get trained, evaluated, registered and deployed, so it holds the training data, the credentials to the data stores and the right to publish a model. Harden it like CI: pinned dependencies, no secrets reachable from untrusted jobs, and a record of every model that shipped.