AI from first principles
A self-paced path from Python basics to training your own models, through the best free material from Harvard, Stanford and practitioners like Andrej Karpathy. Every phase ends with something you build yourself, without a tutorial.
- Lessons
- 37
- Core study time
- about 408 hours
- Level
- Beginner to advanced
Free. Sign in with your email to track progress.
Using AI isn't the same as understanding it
Right now almost everyone is trying AI apps and sharing what they made with them. That's a good start, and it's where most people stop.
If you want a future in AI, using the apps isn't enough. Everyone else can use them too, so the advantage fades fast. What lasts is knowing what's underneath: the maths, the code, how a model is trained and why it behaves the way it does.
This course takes you there, all the way to training a model of your own. And if you only get partway, the attempt will still teach you more than any app can.
The path
- 1
Foundations
8 weeks, 14 lessons
Write Python comfortably and understand the math ML is built on: vectors, matrices, derivatives, probability.
- 2
Classic Machine Learning
2 months, 6 lessons
Understand how models learn from data, how to evaluate them, and how to avoid fooling yourself.
- 3
Deep Learning
3 months, 5 lessons
Build neural networks from scratch, then train a small GPT on your own data.
- 4
Modern AI Systems
3–4 months, 7 lessons
Understand transformers and LLMs in depth; fine-tune, evaluate and deploy real models.
- 5
Specialize & Build in Public
Ongoing, 5 lessons
Pick a niche where you have an edge, contribute to open source, and build credibility toward your company or project.