Sign in
Free Advanced Deep learning

Neural Networks - Zero to Hero

Build a language model from scratch, line by line.

Andrej Karpathy builds neural networks from nothing, in code, on video. The series starts with backpropagation on scalars and ends with a working GPT, each step written out line by line rather than imported from a library.

with Andrej Karpathy
Go to course on Andrej Karpathy

Listed by AI Review Rating. We are not affiliated with Andrej Karpathy and earn nothing from this link.

Neural Networks: Zero to Hero

About Andrej Karpathy

Andrej Karpathy was a founding member of OpenAI and led AI at Tesla. He publishes his teaching free on YouTube with companion notebooks, building each system from scratch on video.

Open the course

Course overview

It is the course people recommend when someone says they understand the diagrams but not the mechanism. Expect to pause, type it yourself, and rerun it until it clicks.

What you will learn

  • Derive and implement backpropagation from first principles
  • Build a character-level language model and improve it step by step
  • Understand what each layer of a transformer is actually doing
  • Write a small GPT end to end without reaching for a framework

Who this course is for

Engineers who want the mechanism you can code and are done with black boxes
Students of ML the companion to a theory course that makes it concrete

Course details

FormatVideo lectures with companion notebooks
LevelAdvanced
PriceFree
Time20 hours
CertificateNo certificate

Learner ratings

No ratings yet. We publish an average once 3 members have rated it, so one opinion never reads as a score.

Taken this course? Sign in or create a free account to rate it. Ratings are members-only so one person cannot set the average.

More courses like this