Glossary
The set of practices and tooling for reliably deploying, monitoring, and maintaining machine learning models in production — applying the same operational rigor DevOps applies to traditional software to the specific lifecycle of ML models.
Why It Matters
A model that performs well in a notebook can degrade silently in production as real-world data shifts away from what it was trained on. MLOps is what catches that — versioning, monitoring, and retraining pipelines that keep a model's real-world performance visible instead of assumed.
This is part of our AI Systems capability — the infrastructure that lets intelligence get embedded into a workflow, not bolted on as a standalone feature.
See the full AI Systems capabilityIs MLOps only relevant for teams training their own models?
Mostly, yes — teams calling a third-party model API (rather than training their own) need less of this, though monitoring for output quality and drift still matters regardless of who trained the underlying model.
What happens without MLOps discipline?
Models silently degrade as real-world data drifts from training data, with no automated signal that performance has dropped — problems usually get discovered from user complaints instead of monitoring.