This course provides a comprehensive introduction to Machine Learning Operations (MLOps), focusing on the reliable and efficient deployment and maintenance of ML models in production. It covers model development, deployment, monitoring, and optimization, and offers hands-on experience with popular MLOps tools and frameworks for managing scalable ML workflows. This course is also offered as an elective for Degree level students.

Lectures-Tutorials-Practicals-Credits: 4-0-0-4

Week 1: Introduction to MLOps; MLOps vs DevOps; Challenges in production ML; MLOps lifecycle and tool ecosystem.

Week 2: ML pipelines and data management; Data engineering practices; Pipeline automation; DVC for data/version control.

Week 3: Feature stores; Role in ML and generative AI; Tools like Feast; Benefits for production systems.

Week 4: CI/CD for ML; Version control with Git; Automated testing and validation; Model delivery strategies.

Week 5: Model development workflows; Experiment tracking (MLflow); Hyperparameter tuning; Small vs large models (LLMs).

Week 6: Model deployment; Containerization and orchestration; Cloud/edge/on-prem deployment; Model serving strategies.

Week 7: Monitoring ML systems; Logging and error tracking; Performance optimization and scaling.

Week 8: ML security; MLSecOps fundamentals; Security risks and tooling.

Week 9: ML governance; Explainability; Bias detection; Ethical AI considerations.

Week 10: MLOps for LLMs (Part 1); Model versioning; CI/CD adaptations for large models.

Week 11: MLOps for LLMs (Part 2); Performance vs cost trade-offs; Observability; Governance (bias, toxicity, explainability).

Week 12: Advanced topics; Federated learning; Edge inference; End-to-end MLOps integration.