Course Syllabus

Module 1: Introduction to MLOps and Python Basics (3 hours)

This module covers the MLOps fundamentals and the Python environment. Learners will understand the ML lifecycle and why automated pipelines are critical. Topics include differences between data science, ML, and MLOps; stages of a production ML workflow; and an overview of MLOps benefits (reproducibility, versioning, CI/CD). The module also reviews the Python tooling (virtual environments, package management, and ML libraries) and sets up development environments (e.g. Jupyter notebooks, Git).

• Subtopics: ML lifecycle overview; Introduction to MLOps concepts; Python setup (venv, IDEs); key Python ML libraries.
• Key features: Hands-on lab: Configure Python environment and Git; Quiz: Basics of MLOps and workflow.

Module 2: DevOps Foundations for MLOps (3 hours)

Students learn the DevOps essentials that underpin MLOps pipelines. This includes version control with Git/GitHub, containerization with Docker, and basic Continuous Integration/Continuous Deployment (CI/CD) principles. Learners build and containerize a simple ML project, and practice setting up automated testing and deployment triggers. These skills are essential for packaging and deploying models reliably.

• Subtopics: Git/GitHub workflows; Docker containers for ML; Introduction to CI/CD pipelines; Continuous integration for model code.
• Key features: Hands-on lab: Containerize a Python model in Docker; Exercise: Automate a test-build pipeline; Quiz: DevOps tools and concepts.

Module 3: Experiment Tracking and Model Registry with MLflow (5 hours)

This module focuses on MLflow, an open-source MLOps platform for tracking experiments and managing models    . Learners install MLflow and use its components: Tracking (logging parameters, metrics, artifacts) and Model Registry (versioning, staging, deployment)    . Using Python notebooks, students run multiple model training experiments, record results in MLflow, and compare them via the MLflow UI. They practice packaging models (MLflow Projects) and registering the best model for deployment.

• Subtopics: MLflow overview    ; Experiment logging (parameters, metrics); MLflow UI dashboard; MLflow Projects; Model packaging and registry.
• Key features: Hands-on lab: Track a classification experiment in MLflow; Project work: Build an MLflow project template; Quiz: MLflow concepts and commands.

Module 4: Pipelines and Orchestration with Kubeflow (5 hours)

In this module, learners explore Kubeflow, an open-source MLOps platform built on Kubernetes    . Students learn to design and run end-to-end ML pipelines using Kubeflow Pipelines (KFP). Starting from simple Python functions or TensorFlow notebooks, they create reusable pipeline components, assemble them into workflows, and execute them in a Kubernetes environment. The module covers pipeline versioning, monitoring runs, and conditional steps. By the end, students can deploy multi-step pipelines (data preprocessing, training, evaluation) on a Kubeflow cluster.

• Subtopics: Kubeflow overview    ; Building pipeline components; Compiling and deploying Kubeflow pipelines; Pipeline versioning and monitoring; Integrating with Kubernetes.
• Key features: Hands-on lab: Develop a Kubeflow pipeline for an end-to-end ML workflow; Exercise: Parameterize and schedule pipeline runs; Quiz: Kubeflow pipeline concepts.

Module 5: Cloud MLOps on AWS SageMaker (5 hours)

This module covers AWS SageMaker and its MLOps tools. Students learn to use SageMaker Studio, SageMaker Pipelines, and Model Registry. The course shows how SageMaker provides “purpose-built tools for machine learning operations (MLOps) to help you automate and standardize processes across the ML lifecycle”.  Learners will create and run SageMaker training jobs, automate them in SageMaker Pipelines, and deploy models to endpoints. The module also demonstrates SageMaker’s integration with MLflow (tracking in SageMaker) and how to monitor model performance.

• Subtopics: SageMaker Studio and Notebooks; SageMaker Pipelines for CI/CD (training→deployment); SageMaker Model Registry; Automating data processing and retraining; Integration with MLflow.
• Key features: Hands-on lab: Build and run a SageMaker training pipeline; Project work: Deploy a model endpoint on SageMaker and perform A/B testing; Quiz: SageMaker MLOps services.

Module 6: Cloud MLOps on Azure Machine Learning (5 hours)

Learners explore Azure Machine Learning and its MLOps capabilities. Topics include Azure ML Workspaces, Pipeline jobs, and model registries. The module emphasizes Azure’s CI/CD integration: students create automated pipelines using Azure DevOps or GitHub Actions to train and deploy models, reflecting how Azure “automate[s] ML workflows using built-in interoperability with Azure DevOps and GitHub Actions”. Students also learn experiment tracking and model versioning in Azure ML, noting Azure’s compatibility with MLflow for experiment logging. By module end, learners can deploy models as Azure endpoints and set up monitoring (data drift, performance).

• Subtopics: Azure ML Workspace setup; Azure ML Pipelines for training/deployment; MLflow tracking in Azure ML; Continuous Deployment with Azure DevOps; Model monitoring (data drift, feedback).
• Key features: Hands-on lab: Execute an Azure ML pipeline to train and register a model; Exercise: Integrate a GitHub Actions CI workflow; Quiz: Azure ML MLOps concepts.

Module 7: Model Monitoring, Governance and Best Practices (3 hours)

This module covers post-deployment MLOps tasks. Students learn how to monitor models in production for accuracy, performance, and data drift. It reviews logging practices and observability (using tools like Prometheus or Azure ML’s monitoring). The course also discusses model governance: version control, lineage tracking, and compliance (e.g. documenting data sources, setting policies). Emphasis is placed on “efficiently monitoring model performance and managing governance” to keep models reliable.

• Subtopics: Production monitoring (accuracy, drift, alerts); Logging and observability for ML; Model versioning and lineage; Compliance and security best practices.
• Key features: Hands-on lab: Set up a monitoring dashboard (e.g. in Azure Monitor or AWS CloudWatch) for deployed models; Exercise: Define governance workflows (model approval, rollback); Quiz: Monitoring and governance strategies.

Module 8: Domain-Specific MLOps Projects (8 hours)

Learners apply MLOps in three industry domains through guided projects. Each project uses the tools learned and follows the full pipeline from data to deployment.

• Project (Finance, 2h): Build an MLOps pipeline for credit risk or fraud detection using synthetic or open financial data. Employ MLflow for tracking experiments and deploy the final model on AWS or Azure with continuous retraining. This reflects how MLOps models aid “risk management and fraud detection” by processing large data streams    .
• Project (Healthcare, 2h): Develop an MLOps workflow for a healthcare prediction (e.g. patient readmission or outbreak forecasting). Use Kubeflow or MLflow to version datasets and models, and containerize the model. This reinforces MLOps in “predictive analytics” where models can anticipate disease patterns    .
• Project (Manufacturing, 2h): Create an MLOps solution for predictive maintenance (e.g. anomaly detection on sensor data). Set up a Kubeflow pipeline or SageMaker pipeline to preprocess IoT data, train a model, and monitor it. This mirrors real-world manufacturing use cases (quality control, predictive upkeep)    .
• Project (Cross-domain Capstone, 2h): In teams, students define a mini-project (choosing any domain dataset) and implement an end-to-end MLOps pipeline incorporating tools from the course (e.g. containerize, track with MLflow, orchestrate with Kubeflow/Azure Pipelines).
• Key features: Each project includes Hands-on labs with Jupyter/Python code, Team exercises simulating a real client requirement, and Quizzes checking understanding of domain-specific considerations.

Module 9: Capstone Review and Certification Prep (3 hours)

The final module consolidates learning and prepares students for certification. It includes a capstone review of all modules and a full mock-project demonstration (from data versioning to monitored deployment). Students also take a practice exam or quizzes modeled on certification objectives (e.g. AWS Certified Machine Learning – Specialty, Azure AI Engineer) to ensure readiness. This wrap-up reinforces best practices and clarifies any gaps.