AI NEWS PAGE DOSSIER

Jeremy Howard

Jeremy Howard's file: Kaggle, medical AI, fast.ai, ULMFiT, practical deep-learning education, and a career spent trying to put powerful models in more hands.

Howard moved from startups and competitive machine learning into medical AI, then co-founded fast.ai around a stubborn idea: deep learning should be usable by ordinary programmers. He helped popularize practical transfer learning, co-created ULMFiT, and later started Answer.AI to keep pushing applied AI outside the giant labs.

HOW THE HELL DID JEREMY HOWARD GET HERE?

  1. Oct 7, 2016

    HE QUITS THE CEO JOB AND STARTS TEACHING DEEP LEARNING

    After leading Enlitic and earlier serving as president and chief scientist of Kaggle, Howard launched fast.ai with Rachel Thomas to make deep learning more accessible.

    RECEIPTS: fast.aifirst-party

  2. Jan 17, 2017

    THE COURSE PUSHES CODERS TOWARD THE RESEARCH FRONTIER

    fast.ai expanded Practical Deep Learning for Coders with a second course focused on implementing recent research and production-grade systems.

    RECEIPTS: fast.aifirst-party

  3. Sep 8, 2017

    FAST.AI MOVES ITS STACK TO PYTORCH

    Howard announced that future fast.ai courses would be built largely around PyTorch and a higher-level library the team had created on top of it.

    RECEIPTS: fast.aifirst-party

  4. May 15, 2018

    ULMFIT MAKES TRANSFER LEARNING WORK FOR LANGUAGE

    Howard and Sebastian Ruder introduced ULMFiT, a transfer-learning method that helped establish fine-tuning as a powerful approach for natural-language tasks.

    RECEIPTS: fast.aifirst-party

  5. 2020

    THE COURSE BECOMES A BOOK-AND-CODE SYSTEM

    fast.ai released a new Practical Deep Learning for Coders course alongside a freely accessible notebook-based book and updated software stack.

    RECEIPTS: fast.aifirst-party

  6. Dec 12, 2023

    TEN MILLION DOLLARS FOR A SMALL APPLIED-AI LAB

    Howard announced Answer.AI, a new research-and-development lab backed by a $10 million investment and focused on making AI useful to more people.

    RECEIPTS: fast.aifirst-party

MORE DOSSIERS