·
Builders & Technical Leaders
Jeremy Howard
Founding CEO |Answer.AI
Co-founder of fast.ai, creator of ULMFiT (precursor to modern LLMs), and founding CEO of Answer.AI. Pioneered transfer learning for NLP and democratized deep learning education.
1papers
GitHubjph00Research Papers
1 paperBiography
Deep Research
Jeremy Howard is an Australian data scientist, entrepreneur, and educator. He is the founding CEO of Answer.AI, an AI R&D lab creating practical end-user products from foundational research breakthroughs, and co-founder of fast.ai, a research institute dedicated to making deep learning accessible. Howard created ULMFiT (Universal Language Model Fine-tuning), the transfer-learning method credited as the precursor to all modern large language models including GPT and Gemini. He previously served as President and Chief Scientist of Kaggle (where he was the top-ranked ML competition participant globally in 2010-2011), founded Enlitic (the first company to apply deep learning to medicine, named one of MIT Tech Review's 50 smartest companies), and founded FastMail and Optimal Decisions Group (acquired by LexisNexis). He spent eight years in management consulting at McKinsey & Co and AT Kearney. Howard is an Honorary Professor at the University of Queensland, a Digital Fellow at Stanford University, and co-authored 'Deep Learning for Coders with Fastai and PyTorch'. His TED talk on the implications of machine learning has over 2.5 million views. He also co-founded the Masks4All movement during COVID-19 and led the largest masks evidence review published in PNAS.
Transfer LearningFine-tuning LLMsDemocratizing Deep LearningNLP & Language ModelsMedical AIAI EducationOpen Source AIPractical AI ProductsWeb Development with PythonDialogue Engineering
Timeline
17 Research17 total
2025
2024
2023
2023-10Research
Appeared on Latent Space podcast: The End of Finetuning2020
2019
2018
2016
2012
2012-01Research
Became President and Chief Scientist of Kaggle2010
Key Contributions
ULMFiT
Universal Language Model Fine-tuning for Text Classification (ACL 2018, with Sebastian Ruder). Pioneered transfer learning in NLP, reducing error by 18-24% on six benchmarks and demonstrating that fine-tuning pre-trained language models is effective for downstream tasks -- the foundation of all modern LLMs.
fastai
One of the world's most popular deep learning frameworks, providing a layered API from high-level learners to low-level callbacks. Powers the free 'Practical Deep Learning for Coders' MOOC taken by hundreds of thousands of students worldwide.
fast.ai Courses
Free deep learning courses co-created with Rachel Thomas that democratized AI education, making cutting-edge techniques accessible without requiring a PhD or expensive compute.
FSDP + QLoRA
Collaboration with Tim Dettmers and Hugging Face enabling training of 70B-parameter language models on consumer GPUs with two RTX 3090/4090 cards, dramatically lowering the barrier to fine-tuning large models.
FastHTML
Python web framework from Answer.AI built on Starlette, Uvicorn, and HTMX, enabling complete web applications in a single Python file for rapid AI product prototyping.
Claudette
Pythonic wrapper around the Anthropic SDK with progressive abstractions, culminating in a complete agentic tool loop in a dozen lines of code.
nbdev
Literate programming system for Python that combines code, tests, and documentation in Jupyter notebooks, generating library code, docs, and CI automatically.
Enlitic
First company to apply deep learning to medical diagnostics (founded 2014). Named one of MIT Tech Review's 50 smartest companies two years running.
Notable Quotes
“
We should treat it as a continuum, and we should have much higher expectations of how much you can do.
“
You can create a complete web application in a single Python file... working with web foundations directly, but entirely in pure Python.
“
We don't have any managers. We don't have any hierarchy from that point of view.
“
The only people who could wield this tool were a very small group of basically, white men with technical skills attached to a small number of esteemed universities. We felt this was a shame.
12 sources(click to expand)
fast.ai -- AboutJeremy Howard (entrepreneur) -- WikipediaULMFiT paper (arXiv)Answer.AI launch announcementFSDP + QLoRA -- Answer.AIAI Magic: Shipping 1000s of products -- Latent SpaceThe End of Finetuning -- Latent SpaceJeremy Howard on Lex Fridman PodcastJeremy Howard on MAD Podcast with Matt TurckSE Radio 391: Jeremy Howard on Deep Learning and fast.aiGitHub profile: jph00Jeremy Howard -- Stanford Digital Economy Lab
Research generated March 19, 2026
Builders & Technical Leaders/Jeremy Howard
All Profiles