Curriculum
The Machine Learning Zoomcamp covers nine main modules plus project work. Each module has video lectures, reading material, and a homework assignment.
For the canonical curriculum (videos, code, exact homework questions), see the GitHub repository.
Modules
- Setup, basic Python and pandas, basic linear algebra refresher.
- Predicting car prices.
- Feature engineering, training, evaluation.
Module 3: Binary Classification
- Customer churn prediction.
- ROC curves, precision-recall, cross-validation.
- Often the most challenging module conceptually.
- FastAPI for serving models.
- Docker for packaging.
- Decision trees, random forests, gradient boosting.
- Loan default prediction.
Module 7: Midterm Project
- Three weeks of independent work.
- Apply what you have learned through Module 6.
Module 8: Neural Networks and Deep Learning
- PyTorch fundamentals.
Module 9: Serverless Deployment
- AWS Lambda with ONNX runtime.
- Deploying ML services on Kubernetes.
- Optional and potentially outdated. Focus on Kubernetes (Module 10) instead.
Capstone Project
- Three weeks of independent work after Module 10.
Learning philosophy
The course follows a project-first approach: theory is introduced as needed to solve real problems. Practical over theoretical, focused on ML engineering skills used in industry.
The course is community-driven. Learning together, sharing your progress, and helping each other are part of the design.
Pace
A typical week:
- Watch the module videos (3 to 5 hours).
- Code along and experiment (3 to 5 hours).
- Complete the homework (2 to 5 hours).
Plan for 10 to 15 hours per week. Module 4 (model evaluation) is harder; allocate more time.
For the project blocks (midterm and capstone), plan 2 to 3 weeks of focused work each.
What changes between cohorts
For changes specific to the current cohort, see What’s New.