MLOps Minimum Viable Vocabulary
Recall the operational difference between common MLOps terms.
What is a model artifact? The serialized trained output plus metadata needed to load or register it. It is not the whole system; it is one versioned artifact inside a workflow. What is a training pipeline? A repeatable process that transforms data, trains candidates, evaluates them, and emits artifacts. A notebook can inspire it, but the pipeline must be rerunnable. Training pipeline vs inference service Confusing these creates bad estimates and missing owners. Objection: "Can we just save the model file in the shared drive?" A file is not enough; we need version, data lineage, metrics, owner, and rollback target…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in