Back to glossary

MLOps

The set of practices combining machine learning, DevOps, and data engineering to reliably deploy, monitor, and maintain ML models in production.

MLOps bridges the gap between training a model in a notebook and running it reliably in production. It covers the full lifecycle: data versioning, experiment tracking, model training pipelines, evaluation, deployment, monitoring, and retraining triggers.

The MLOps maturity spectrum ranges from Level 0 (manual everything — Jupyter notebook to production) to Level 3 (fully automated CI/CD for ML — automatic retraining triggered by data drift detection). Most growth teams should aim for Level 1-2: automated training pipelines, version-controlled experiments, automated evaluation against test sets, and basic model monitoring.

Key MLOps tools include experiment trackers (Weights & Biases, MLflow), feature stores (Feast, Tecton), model registries (MLflow, Vertex AI), serving platforms (BentoML, Seldon), and monitoring solutions (Evidently, Arize). For teams using primarily LLMs and APIs, "LLMOps" is an emerging subset focused on prompt management, cost tracking, evaluation pipelines, and guardrails — with tools like LangSmith and Helicone filling this niche.

Related Terms

Further Reading