Course Syllabus

  Module 1: Foundations of AI & ML (with Python) – Introduces AI vs. ML, types of learning (supervised, unsupervised, reinforcement), and common ML tasks. Covers core algorithms (e.g. regression, classification, clustering) at a conceptual level. Defines machine learning as a branch of AI where “computers learn from data and improve with experience without being explicitly programmed”. The module also sets up the Python environment: installing Anaconda/Jupyter, and using libraries like NumPy, Pandas, and scikit-learn. Learners run a simple Python notebook to train a basic model, illustrating how Python’s ecosystem supports ML development. Hands-on: load a sample dataset and execute a complete Python-based ML workflow (data load, train, evaluate).

  Module 2: Data Preparation & Exploratory Analysis – Focuses on data quality and feature engineering. Topics include data cleaning (handling missing values, outliers), transformation (scaling, encoding), and feature selection. Students learn to use Pandas and visualization tools (Matplotlib/Seaborn) to explore datasets, identify patterns, and engineer new features. Emphasizes how “garbage in, garbage out” applies to ML – proper data prep is critical. Hands-on: given a real-world dataset, perform cleaning and feature engineering in Python, documenting steps.

  Module 3: Supervised & Unsupervised Learning Algorithms – Covers algorithmic techniques for prediction and clustering. In supervised learning, students study linear/logistic regression, decision trees, k-nearest neighbors, and support vector machines. In unsupervised learning, topics include k-means clustering and principal component analysis for dimensionality reduction. Each topic includes intuition and mathematical foundations. Hands-on: using scikit-learn, participants train and compare different models on structured data (e.g. classification of iris data, regression on housing prices).

  Module 4: Model Evaluation and Tuning – Teaches how to assess and improve ML models. Topics include train/test splits, cross-validation, and performance metrics (accuracy, precision/recall/F1 for classification; RMSE/R² for regression). Covers overfitting vs. underfitting and bias-variance tradeoff. Introduces hyperparameter tuning techniques (grid search, random search). Hands-on: evaluate trained models from Module 3, plot learning curves, and use cross-validation to select hyperparameters.

  Module 5: Introduction to MLOps Practices – Explains the motivation for MLOps: applying DevOps principles to the ML lifecycle. Learners see that MLOps “sits at the intersection of data science, DevOps, and software engineering” and is focused on reliably deploying and operating ML models at scale. Covers core MLOps activities: version control for data and code, continuous integration/continuous delivery (CI/CD) pipelines for ML, and the concept of experiment tracking and reproducibility. Introduces containerization (Docker) and basics of Kubernetes. Hands-on: set up Git versioning for an ML project and containerize a simple ML service.

  Module 6: Experiment Tracking with MLflow – Dives into MLflow, an open-source ML lifecycle platform. Topics include MLflow’s four components: Tracking (log parameters, metrics, artifacts in experiments), Projects (reproducible run specifications), Models (packaging), and Model Registry (centralized model store). Students learn to log experiments via the MLflow Python API and use the MLflow UI to compare runs. Hands-on: instrument an existing training script to use MLflow Tracking Server; run multiple training experiments and organize them with MLflow’s registry.

  Module 7: Building Pipelines with Kubeflow – Covers workflow orchestration using Kubeflow on Kubernetes. Introduces Kubeflow’s mission: an open-source ecosystem to address every stage of the ML lifecycle on Kubernetes. Topics include Kubeflow Pipelines for defining end-to-end workflows, Notebook servers for development, and KFServing/KServe for model serving. Students learn to define containerized pipeline steps (data processing, training, evaluation) and connect them in Kubeflow Pipelines. Hands-on: deploy a local (or cloud) Kubeflow cluster, create a multi-step pipeline (e.g. data ETL + model train + deployment), and run it.

  Module 8: MLOps on AWS SageMaker – Explores cloud-based MLOps with Amazon SageMaker. Covers SageMaker Studio and Projects, training jobs, and endpoints for deployment. Emphasizes SageMaker’s “purpose-built tools for MLOps” that automate and standardize ML lifecycle processes. Students learn about SageMaker Pipelines (CI/CD for ML), Experiments (tracking), and Model Monitor (quality monitoring). Hands-on: using AWS Free Tier, build a SageMaker pipeline that trains a model on S3 data and deploys it as a real-time endpoint. Explore SageMaker Experiments to track runs.

  Module 9: MLOps on Azure Machine Learning – Focuses on Microsoft Azure’s enterprise ML service. Introduces Azure ML Workspace, Compute, and Designer pipelines. Covers Azure’s MLOps capabilities: managing pipelines with Azure Pipelines, registering models in the Azure Model Registry, and automating deployment to Azure Kubernetes Service. Students learn how Azure implements DevOps for ML (CI/CD, versioned environments, monitoring). Hands-on: set up an Azure ML workspace, run a pipeline that trains a model, register the model, and deploy it as an Azure ML endpoint.

  Module 10: Deployment, Monitoring, and Capstone Project – Covers final steps and integration. Topics include model deployment strategies (batch vs. real-time, container serving), continuous monitoring (data/model drift detection), and ML governance. Emphasizes automating these steps via CI/CD pipelines (e.g. using Jenkins/GitHub Actions or cloud pipelines). The capstone project tasks participants to build an end-to-end ML solution using learned tools: from data prep to deploying a model on the cloud with monitoring. Hands-on: work in teams to complete a real-world ML project (such as image classification or recommendation) and present the results.

Key Features

  Hands-on Labs: Each module includes practical labs using real datasets. Students will code in Python notebooks to build and train models, implement pipelines, and deploy services. This reinforces theory with experience.

  Project-based Learning: The course culminates in one or more real-world projects (for example, deploying an image classification or customer-churn model) to simulate production scenarios.

  Industry Tools Integration: Extensive use of popular MLOps platforms – MLflow for experiment tracking, Kubeflow for pipeline orchestration, and cloud services AWS SageMaker and Azure ML for scalable training and deployment. Students gain familiarity with tools actually used in enterprise ML workflows.

  CI/CD and Automation: Coverage of best practices like continuous integration and delivery for ML. For example, AWS notes the importance of “ML CI/CD integration… for faster time to production” and continuous monitoring to maintain model quality. Labs may include setting up automated pipelines (e.g. using Git, Docker, Jenkins/GitHub Actions, or cloud DevOps tools).

  Python-focused Environment: Throughout the course, Python is the primary language. Python’s widespread use in ML (with libraries like TensorFlow, PyTorch, scikit-learn, Keras) ensures that learners build transferable skills.

  Certification Preparation: The curriculum aligns with skills needed for industry certifications (e.g. AWS Certified Machine Learning – Specialty, Microsoft Certified: Azure AI Engineer). Practice exams and study guides can be provided to encourage certification.

  Comprehensive Support: Instructors review best practices in ML experimentation, reproducibility, and governance. Discussions on topics like model fairness and ethics may be included, depending on audience interest. A certificate of completion is awarded to verify the hands-on experience gained.

.