Jeremy Howard
Deep Learning Educator and fast.ai Co-Founder
Jeremy Howard is an Australian AI researcher and entrepreneur who co-founded fast.ai, a nonprofit that produces free deep learning courses and maintains the fastai software library. He previously co-founded Enlitic, an early medical AI company, and led Kaggle as its president. He does not hold a PhD in machine learning or a related field, having studied philosophy at the University of Melbourne, though his practical contributions to the field through software and education are substantial.
The fast.ai "Practical Deep Learning for Coders" course, which he teaches along with Rachel Thomas, takes a top-down pedagogical approach — starting with working code and practical applications before introducing underlying mathematics and theory. This approach contrasts with the bottom-up methodology typical of formal machine learning curricula. The course is free and has reached hundreds of thousands of students. Many participants have subsequently published research, entered the field professionally, or built companies, making fast.ai one of the more measurably impactful free AI education resources available.
The fastai library, built on PyTorch, provides high-level abstractions for training neural network models. It has been used extensively in academic research, including at international machine learning competitions, and enabled some researchers without large computational resources to achieve competitive results. The library's design choices have been subjects of discussion in the deep learning community, including trade-offs between abstraction convenience and understanding of underlying operations — a reflection of the pedagogical philosophy that also shapes the courses.
Howard has been publicly vocal on AI policy, safety, and the social implications of rapid AI development. His positions have included criticism of the concentration of AI capabilities in a small number of large technology companies and support for regulation and broader access to AI tools. These views are consistent with his education-democratization work, though they place him in a contested space where different stakeholders have conflicting interests in how AI development proceeds.