Everything to learn, in order.
Sixteen modules in three parts. The first two lessons of every module are free.
How models work
Classical ML to PyTorch, transformers and fine-tuning, with small numbers you can check by hand.
Python and maths toolkit
Vectors, probability and gradients, with NumPy.
02Classical machine learning
Regression, trees, boosting, and how to tell if a model is any good.
03Neural networks
From one neuron to backprop, by hand, with small numbers.
04Deep learning in PyTorch
Tensors, autograd and training loops: the tools real ML work uses.
05Tokens and embeddings
How text becomes numbers a model can work with.
06Transformers
Attention, the full architecture, and how the next token is chosen.
07Training and fine-tuning
Pretraining, LoRA, preference tuning and the data behind them.
Build with LLMs
Prompting, retrieval, agents, evaluation and what it takes to run them in production.
Prompting and context
Prompts, structured outputs, tools, and what to put in the window.
09Search and retrieval
RAG from chunking to reranking, and how to evaluate it.
10Agents
The agent loop, memory, tools, and how agents fail.
11Evaluation
Decide what good means, then measure it without fooling yourself.
12Inference and serving
Where the time goes, and how to make models fast and cheap.
13Production AI and MLOps
Routing, fallbacks, budgets, monitoring and safety in real systems.
Design, data and code
Design whole ML systems, analyse experiments honestly, and write core algorithms from scratch.