Doug Does · Three courses

Finish a machine learning project you can explain

Each course ends with something concrete: a trained model, a measured comparison, or worked examples you changed yourself. Every lesson is open without an account. Start where your background fits.

01 / 6 LESSONS

Train a small language model on Shakespeare

A downloadable Python project that trains a small character-level transformer on your computer, with tests, data, and supplied trained models to compare against.

Basic Python · Explore the course →
02 / 5 LESSONS

Fine-tune a model to answer questions with SQL

A before-and-after comparison on WikiSQL questions the model never trained on, showing which answers improved and which didn't. The optional lab runs on Apple Silicon.

Basic Python + model basics · Explore the course →
03 / 6 LESSONS

Learn how models predict and improve

Browser exercises where you change a model's numbers and see the effect on predictions and error, plus optional Python exercises if you know basic Python.

Basic algebra · Explore the course →

Find your way in

Six ways to learn AI · Karpathy study companion

ML reference slides · Python examples