Using AI tools is easy now. Understanding how they work well enough to build one is a different skill, and it's the one I want.
So I'm starting from the beginning: Python, then the maths underneath machine learning, then classic ML, deep learning, and finally modern language models. The goal at the end is to train and ship models of my own, and eventually build an AI company or an open-source project around them.
The plan
The path has five phases:
- Foundations. Python, NumPy, and a refresh of algebra, statistics, linear algebra and calculus.
- Classic machine learning. How models learn from data, and how to evaluate them honestly.
- Deep learning. Neural networks from scratch, ending with a small GPT trained on my own data.
- Modern AI systems. Transformers, fine-tuning, retrieval and deployment.
- Specialise and build in public. Pick a niche, contribute to open source, and talk to real users.
Almost everything on it is free: Harvard's CS50P, 3Blue1Brown, Andrew Ng, Stanford's lectures, fast.ai, and Andrej Karpathy's Zero to Hero series.
One rule
Every phase ends with something built without a tutorial open. Watching a lecture feels like learning; rebuilding it from a blank file is the real test.
Why in public
Writing forces clarity, and a trail of projects on GitHub is worth more than any certificate. If you want to follow along, the full course is on this site, free, with progress tracking. You can start at the first lesson today.