Automated ML pipeline development (CI/CD/CT), model registry and version control, model deployment and containerisation on AWS/Azure/GCP, production performance monitoring, data drift detection, automated retraining pipelines, compute cost optimisation, and LLMOps for Generative AI applications — including prompt versioning, hallucination monitoring, RAG retrieval quality tracking, and token cost management.

MLflow for experiment tracking and model registry, Kubeflow for Kubernetes-native pipeline orchestration, AWS SageMaker Pipelines and Step Functions for AWS environments, Azure ML Pipelines for Azure, Evidently AI for drift detection, Langsmith and Langfuse for LLMOps tracing and evaluation, GitHub Actions for CI/CD, and Terraform and Kubernetes for infrastructure as code.

LLMOps extends MLOps practices for Large Language Model applications. Traditional MLOps monitors prediction accuracy and data drift. LLMOps adds: prompt versioning and evaluation, hallucination monitoring to detect factually incorrect LLM outputs, RAG retrieval quality monitoring, token cost management for usage-billed LLM APIs, and content safety monitoring for prompt injection attempts and policy violations.