Roadmap
Machine learning
The maths, the classic models and deep learning, from the people who teach it best.
Who it's for
Programmers comfortable with Python who want to understand and build models, not just call them.
Where it leads
- Machine learning engineer
- Data scientist
- Applied scientist
- AI engineer
By the end you can
- Explain how a model learns and how it fails
- Train and evaluate models with scikit-learn
- Build and train a neural network from scratch
- Take a model from notebook to a deployed service
- 1
The maths you actually need
Linear algebra, calculus and probability at the level ML uses them, intuition first.
- 2
Core machine learning
Regression, classification, overfitting and evaluation. Everything else builds on these ideas.
- 3
Hands-on with scikit-learn
Pipelines, cross-validation and real datasets, the way models get built at work.
Do this: Enter a Kaggle playground competition and beat your own first submission three times.
- 4
Neural networks from scratch
Build backpropagation yourself once and deep learning stops being magic.
Do this: Follow micrograd end to end and train a tiny network on a dataset of your own.
- 5
Deep learning in practice
Train real models on images and text with PyTorch, and learn what makes them work.
- 6
Ship it
A model in a notebook helps nobody. Learn to version, deploy and monitor it.
Do this: Serve your best model behind an API with input validation and basic monitoring.
Learning this with others is easier.
Ask questions when you're stuck, find people on the same path, and track your progress on DevLearn.