Colin Raffel

Career

Colin Raffel completed a PhD in electrical engineering at Columbia University in 2016, advised by Daniel P. W. Ellis in a laboratory for speech and audio research. Raffel then joined Google Brain in 2016 and stayed until 2021, holding research scientist roles from 2017 and becoming a staff research scientist in 2020.²

Raffel was an assistant professor of computer science at UNC Chapel Hill from 2020 to 2023 and has been a faculty researcher at Hugging Face since 2021. Since 2023 Raffel has been an associate professor at the University of Toronto and an associate research director at the Vector Institute.²¹

Research

Raffel's lab works on decentralizing, democratizing and de-risking large-scale AI. At Google, Raffel was first author of the T5 paper, which recasts text-based language tasks as text-to-text problems and introduced the Colossal Clean Crawled Corpus.¹³

Raffel's later papers cover training data, model merging and collaborative model development. They include "Scaling Data-Constrained Language Models" (2023), TIES-Merging for combining fine-tuned models (2023), Git-Theta, a git extension for machine learning models (2023), and the open datasets FineWeb (2024) and The Common Pile v0.1 of public domain and openly licensed text (2025).²

Work and contributions

• T5 (Text-to-Text Transfer Transformer) Open-source code

  First author of the T5 paper (2019), which introduces a unified text-to-text framework and the Colossal Clean Crawled Corpus; the Apache-2.0 google-research/text-to-text-transfer-transformer repository links the paper and lists Raffel first in its citation.³⁴

Timeline

1. 2016Completed a PhD in electrical engineering at Columbia University and joined Google Brain.²
2. 2017Became a research scientist at Google Brain.²
3. Oct 2019Posted the T5 paper on unified text-to-text transfer learning as first author.³
4. 2020Became an assistant professor of computer science at UNC Chapel Hill and a staff research scientist at Google Brain.²
5. 2021Became a faculty researcher at Hugging Face.²
6. 2023Became an associate professor at the University of Toronto and associate research director at the Vector Institute.²
7. 2025Co-authored The Common Pile v0.1, a dataset of public domain and openly licensed text.²

Other documented roles

Roles the sources document that are not confirmed as current.

• Research Scientist, Senior Research Scientist, and Staff Research Scientist (2017–2021), Google Brain²⁵

Honors and recognition

• Best Student Paper, ISMIR 2015 · International Society for Music Information Retrieval, 2015²
• NSF CAREER Award · U.S. National Science Foundation, 2022²
• NeurIPS Outstanding Paper (Runner-Up) for "Scaling Data-Constrained Language Models" · Conference on Neural Information Processing Systems, 2023²
• Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies (Runner-Up) · Privacy Enhancing Technologies Symposium, 2023²
• ICBS Frontiers of Science Award in Theoretical Computer and Information Sciences · International Congress of Basic Science, 2024²

Where they work

• University of Toronto 3 people in the wiki
• Vector Institute 4 people in the wiki
• Hugging Face 34 people in the wiki

Related people

• Geoffrey HintonAlso at University of Toronto
• Sanja FidlerAlso at University of Toronto
• Arthur MenschLanguage models · Training systems
• Luke ZettlemoyerLanguage models · Training systems
• Jeffrey QuesnelleLanguage models · Training systems
• Jakub PachockiLanguage models · Training systems
• Liliang RenLanguage models · Training systems
• Yonghui WuLanguage models · Training systems
• David GrangierLanguage models · Training systems
• David D. CoxLanguage models · Training systems