Machine Learning Operations

Module No.	Detailed Content	Hrs (26)	CO
1	Introduction to MLOps: CI/CD Pipelines, MLOps lifecycle, Versioning & Reproducibility-Reproducibility using Git, Environment management with Conda, Basics of version control,Structure ML projects for maintainability	04	CO1
2	CI/CD Automation with GitHub Actions & Jenkins for ML Pipelines: Continuous integration and deployment in MLOps workflows, Automate ML pipelines using GitHub Actions and Jenkins for model training and testing	05	CO2
3	API Development and Serving Models with FastAPI + Docker: Build RESTful APIs for ML models using FastAPI and containerize them with Docker, Deploying inference services locally and online	04	CO3
4	Experiment Tracking and Model Management with MLflow & DVC: Tools for tracking experiments (MLflow) and managing datasets/models (DVC), Explore logging parameters, metrics, and visualizing model runs across versions.	03	CO4
5	Cloud Deployment using Azure ML & Google Vertex AI: Deploying ML models to production on cloud platforms - Model registry, endpoints, container deployment, and comparison of Azure ML and GCP Vertex AI features.	05	CO5
6	End-to-End Cloud-Integrated MLOps Pipeline: Workflow from versioning and model development to cloud-based CI/CD deployment	05	CO6

Sr. No.	Title of the Experiment	CO
1	Set up ML project structure with Git and Conda. Offline: Git init, Conda, requirements. Online: Azure ML Environments.	CO1
2	Automate training and testing using GitHub Actions. Offline: YAML CI on localhost. Online: GitHub Actions to Azure ML pipeline.	CO1, CO2
3	Jenkins job to automate model training. Offline: Jenkins local job with scripts. Online: Trigger training on GCP Vertex AI.	CO2
4	FastAPI-based ML inference API. Offline: FastAPI with Uvicorn. Online: Deploy to Azure App Services or GCP Cloud Run.	CO3
5	Containerize ML API using Docker. Offline: Build and test locally. Online: Push to Azure Container Registry or GCP Artifact Registry and deploy on AKS or GKE.	CO3, CO5
6	Track experiments using MLflow. Offline: MLflow UI on localhost. Online: MLflow integration with Azure ML.	CO4
7	Manage data and models using DVC. Offline: DVC init and local tracking. Online: Remote storage on Azure or GCP.	CO4
8	Evaluate and register models using MLflow Registry. Offline: Local SQLite-based registry. Online: Azure ML or Vertex AI registry.	CO4, CO5
9	Deploy model as REST API using Docker. Offline: Docker-based local deployment. Online: Azure Online Endpoints or GCP AI Endpoints.	CO3, CO5
10	End-to-End MLOps pipeline with CI/CD, versioning, deployment, and visualization. Offline: Fully local pipeline. Online: Azure ML or GCP CI/CD setup.	CO6

Sr. No.	Title	Tools
1	Lab 1 – ML Project Setup & Reproducibility	Git, GitHub, Conda/virtualenv, Google Colab
2	Lab 2 – CI/CD with GitHub Actions	GitHub Actions, Python, scikit-learn
3	Lab 3 – Jenkins for Local Automation	Jenkins (local), Python
4	Lab 4 – FastAPI for ML Inference	FastAPI, Uvicorn, scikit-learn
5	Lab 5 – Dockerize ML API	Docker Hub / Desktop or Azure
6	Lab 6 – Experiment Tracking with MLflow	MLflow, scikit-learn
7	Lab 7 – Data & Model Versioning with DVC	DVC, GitHub
8	Lab 8 – Capstone Project	GitHub Actions, FastAPI, Docker, MLflow, DVC, Heroku (free tier)

Tools: GitHub, Colab, Docker Desktop, Heroku free tier, MLflow, DVC, FastAPI.
Cloud services like Azure ML/Vertex AI may be replaced with Heroku/GitHub Actions/DockerHub.