What is 'MLOps'?

MLOps addresses the gap between ML model development (data science team) and production deployment/maintenance (operations team). The key components are version control for data and models (MLflow, DVC), automated training pipelines (Kubeflow, SageMaker), model deployment (REST APIs, batch inference), model monitoring (drift detection, performance degradation), and automated retraining triggers. In banking, a fraud detection model must be continuously monitored. Fraud patterns evolve rapidly, and models need retraining as new attack patterns emerge.