fast.ai is a research and education outfit founded in 2016 by Jeremy Howard and Rachel Thomas with the goal of making deep learning accessible beyond specialist research labs. Both founders had backgrounds applying machine learning in industry before turning to teaching, and Howard has since gone on to co-found Answer.AI, continuing fast.ai's research and open-source work alongside it.
The core offering is 'Practical Deep Learning for Coders,' a project-based course that teaches neural networks by having students train working models from the first lesson, deferring heavy math and theory until after students have working intuition. That teaching approach is backed by the fastai Python library, an open-source layer on top of PyTorch that reduces the code needed to train competitive models, and which the courses use directly.
fast.ai is built for programmers who already know Python and want to apply deep learning practically, without first completing a formal machine learning degree. Courses and course notes are entirely free, released as recorded video and open notebooks; the only ongoing cost for learners is optional cloud GPU time on platforms like Kaggle, Colab, or Paperspace, which fast.ai does not charge for itself.
