Get verified
All guests

Jeremy Howard

Not on Prends yet

Co-founder of fast.ai and founding CEO of Answer.AI

Speaks English6 podcast appearances

dialogue engineering · AI-assisted coding · continued pre-training

About

Jeremy Howard is a data scientist, researcher and educator who co-founded fast.ai, a research institute focused on making deep learning accessible. He is the founding CEO of Answer.AI, an R&D lab that turns research advances into practical products, including the Solveit learning platform, the FastHTML Python web framework and the ModernBERT encoder models.

Known for

  • Created ULMFiT, an early transfer learning method for language models that underpins the pre-train then fine-tune approach
  • Co-founded fast.ai and built the fastai deep learning library, and co-wrote the book Deep Learning for Coders with Fastai and PyTorch
  • Served as President and Chief Scientist of Kaggle, after ranking first among competitors there two years running
  • Was founding CEO of Enlitic, an early company applying deep learning to medicine
  • Founded Answer.AI, whose releases include ModernBERT, FastHTML and the Solveit platform

Key points

  • Argues that pre-training and fine-tuning should be treated as one continuous process, gradually mixing in higher quality and task data, rather than as separate stages
  • Argues that spreading AI capability widely through education and accessible tools serves society better than concentrating it in a few institutions
  • Argues that leaning on AI to generate code can give developers an illusion of control while eroding the hands-on understanding they need, and favors interactive, notebook-style work that keeps people learning

Recent appearances

References hosts use to judge fit

Researched from public sources, checked 2026-10-08. This person has not claimed the profile yet.