Effective MLOps
- English
- Professional
- Online : 1250CHF
The training provides detailed explanations and realistic examples of the motivations, objectives, practices, techniques, and tools for implementing and optimizing Machine Learning operations. Hands-on labs enable participants to practice essential MLOps techniques such as environment setup, data versioning, experiment tracking, model deployment, CI/CD automation, monitoring, testing, and responsible AI integration.
Day 1 : Foundations of MLOps and Experiment Management
- Understand the scope of MLOps and how it complements DevOps in managing the ML lifecycle
- Explore and preprocess data with real-world constraints (missing data, class imbalance, etc.)
- Track experiments, models, and versions using DVC and MLflow
- Set up the MLOps environment: Git, Python virtual env, and Docker
- Explore and preprocess the fraud detection dataset (EDA, missing values, class imbalance)
- Train, version, and track models using DVC and MLflow
Day 2 : Modularization, Deployment & CI/CD Pipelines
- Transform ML notebooks into modular, production-grade code with clean architecture
- Build a complete web application for ML inference using FastAPI (backend) and Streamlit (frontend)
- Containerize and deploy services locally and in the cloud with Docker, AWS ECS, and GitHub Actions
- Refactor Jupyter notebooks into modular Python scripts and packages
- Build and connect a REST API (FastAPI) with a frontend (Streamlit)
- Containerize and deploy the full ML app using Docker, Docker Compose, AWS ECS, and GitHub Actions
Day 3 : Testing, Monitoring & Responsible AI
- Implement testing strategies for ML systems (unit, integration, data validation)
- Monitor ML performance in production and handle model/data drift effectively
- Ensure fairness and explainability through Responsible AI practices
- Write unit and integration tests for data and model components
- Set up data and model validation checks using Deepchecks
- Implement monitoring and explainability features on the deployed pipeline
This course is available online and onsite and fully customizable to your needs.
*The course is also available in French.
Theory