PeopleofSuperintelligence

Researcher

Jeff Donahue

Also known as Jeffrey Donahue

Jeff Donahue is a machine learning researcher who co-authored papers on deep convolutional features for computer vision. He completed a computer science degree at UT Austin in 2011 and a doctoral thesis at UC Berkeley in 2017. He worked as an engineer at Pinterest from 2014 to January 2017 and as a research scientist at DeepMind from 2017 to 2025. His co-authored work includes the Caffe deep learning framework, a large-scale GAN image-synthesis paper, and the Flamingo visual language model. The Longuet-Higgins Prize and the ICML Test of Time Award both honored his papers in 2024.

Timeline

  1. 2007–2011

    Honors thesis at UT Austin

    He completes a Turing Scholars degree in computer science at UT Austin in 2011, with an honors thesis supervised by Kristen Grauman.

  2. 2011–2017

    A Berkeley doctorate

    He completes a doctoral thesis at UC Berkeley in 2017, with Trevor Darrell as advisor. The thesis is titled Transferrable Representations for Visual Recognition.

  3. 2014

    DeCAF and Caffe

    He is one of the authors of DeCAF, a paper on deep convolutional activation features for generic visual recognition. He is also among the authors of Caffe, a deep learning framework submitted to arXiv in June 2014.

  4. 2014–2017

    An engineer at Pinterest

    He works as an engineer on Pinterest's Visual Discovery team from 2014 to January 2017.

  5. 2017–2025

    Research scientist at DeepMind

    He works as a research scientist at DeepMind from 2017 to 2025, and the Veo 3 credits name him.

  6. September 2018

    Large-scale GAN training

    He co-writes Large Scale GAN Training for High Fidelity Natural Image Synthesis with Andrew Brock and Karen Simonyan, submitted to arXiv in September 2018.

  7. April 2022

    The Flamingo visual language model

    He is one of the authors of Flamingo: a Visual Language Model for Few-Shot Learning, submitted to arXiv in April 2022.

  8. 2024

    Two awards in 2024

    In 2024, the Longuet-Higgins Prize goes to Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation, which he co-wrote with Ross Girshick, Trevor Darrell and Jitendra Malik. The ICML Test of Time Award honors DeCAF in 2024.

Sources and credits

Photo: We have no published photo of Jeff Donahue yet. Send us a free-licensed photo.

  1. Jeff Donahue | alphaXiv alphaxiv.org. Checked 14 October 2026.
  2. Transferrable Representations for Visual Recognition | EECS at UC Berkeley www2.eecs.berkeley.edu. Checked 14 October 2026.
  3. DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition proceedings.mlr.press. Checked 14 October 2026.
  4. [1408.5093] Caffe: Convolutional Architecture for Fast Feature Embedding arxiv.org. Checked 14 October 2026.
  5. Veo 3.1 — Google DeepMind deepmind.google. Checked 14 October 2026.
  6. [1809.11096] Large Scale GAN Training for High Fidelity Natural Image Synthesis arxiv.org. Checked 14 October 2026.
  7. [2204.14198] Flamingo: a Visual Language Model for Few-Shot Learning arxiv.org. Checked 14 October 2026.
  8. Everything About CVPR 2025 analyticsvidhya.com. Checked 14 October 2026.
  9. ICML Test Of Time DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition icml.cc. Checked 14 October 2026.