Definition
MLOps covers the whole life of a model: data pipelines, training, versioning, testing, deployment, monitoring for drift and retraining. Google's widely cited guide describes maturity levels from manual processes to fully automated continuous training and delivery.
Much enterprise AI still runs on classic machine learning (forecasting, scoring, recommendations), and MLOps remains the backbone for it. LLMOps and AgentOps build on the same habits for generative and agentic systems.