Build an LLM from scratch.
Become an AI Engineer.
Twelve weeks, from a bigram model to your own fine-tuned, deployed GPT. Learn alongside real people. Every lesson handmade by Angela Yu.
No payment now. Waitlist members hear first.
Not another video course. A frontier learning environment to build, explore, and upskill together.
Fully interactive
Every concept is something you run, poke and break in the browser. Write the code, watch the gradients move, see the loss fall.
Study with real people
Your whole cohort moves through the same week at the same time. Compare notes, debug together, and have people who notice if you fall behind.
A voice AI tutor you can talk to
Just ask out loud. Your tutor hears the question, sees your code and the step you’re on, and talks you through it, without handing over the answer.
Lessons handmade by Angela Yu
Every explanation, diagram and exercise is crafted by the instructor who’s taught millions to code. Clear, warm, and never hand-wavy.
This isn’t a slideshow. It’s an interactive video.
Videos you can write code into and touch. Every lesson plays through like a video, but you can stop at any moment to drag, type and run what’s on screen. These three are straight from the course.
Don’t just use an LLM. Build your own.
Every week ends with something you built yourself, running in front of you: an attention map, a training curve, a tokenizer, a chat model. By week twelve, every line of your GPT is yours.
From one bigram to your own deployed GPT.
One build per week, each on top of the last. Tap any week to see what’s inside.
Train your first model: predict the next character from the one before it.
Dr Angela Yu
Angela trained as a doctor and worked in the NHS before turning to code full time. She’s been programming since she was 12, co-founded the London App Brewery bootcamp, and became one of Udemy’s most popular instructors, teaching millions of students through courses like the 100 Days of Code Python bootcamp.
Companies including Twitter, Facebook and Google have invited her to teach their teams. For this course she is handmaking every lesson, so the deepest ideas in AI feel as approachable as your first line of code.
Twelve weeks. One model, built side by side.
A new week unlocks for the whole cohort at once, so you’re always learning the same thing as the people around you.
No fixed class times. Work through the week whenever suits you, with your cohort a message away.
Between check-ins with your cohort, talk problems through with your voice tutor.
Finish with your own trained, fine-tuned and deployed language model.
Be first in line for the founding cohort.
Leave your email and we’ll tell you as soon as enrolment opens. No spam, just the start date.