PeopleofSuperintelligence

Researcher

James Thornton

James Thornton is a research scientist at Google DeepMind whose work on diffusion models connects generative modelling with optimal transport. After earning undergraduate and master's degrees in statistics at the University of Warwick in 2015 and 2016 and working as an analyst at BlackRock from 2016 to 2018, he completed a DPhil in Oxford's Department of Statistics under George Deligiannidis and Arnaud Doucet in 2023. With Valentin De Bortoli, Jeremy Heng and Doucet, he introduced the Diffusion Schrödinger Bridge as a NeurIPS 2021 spotlight paper, and he coauthored Riemannian Score-Based Generative Modelling, which won a NeurIPS 2022 Outstanding Paper Award. From 2023 to 2025, he was a research scientist in Apple's Machine Learning Research group. Since March 2025, he works in the GenMedia team at Google DeepMind on large-scale multimodal generation with flow and diffusion models.

Badges

Timeline

  1. 2012–2016

    Statistics degrees at Warwick

    He studies statistics at the University of Warwick as an undergraduate from 2012 to 2015 and completes an MSc in statistics from 2015 to 2016. At Warwick, he works with Anthony Lee on Monte Carlo methods for Bayesian non-parametric models.

  2. 2016–2019

    Analyst at BlackRock

    From August 2016 to August 2018, Thornton works as an analyst at BlackRock. From October 2018 to October 2019, he is a probationary research student at the University of Oxford.

  3. 2019–2023

    A DPhil in Oxford Statistics

    In October 2019, he starts a DPhil in the Computational Statistics and Machine Learning research group of Oxford's Department of Statistics, supervised by George Deligiannidis and Arnaud Doucet. He completes the degree in August 2023.

  4. 2021

    Differentiable particle filtering

    With Adrien Corenflos, George Deligiannidis and Arnaud Doucet, he introduces a differentiable particle filter built on optimal transport, with convergence results. The paper is accepted as an oral at ICML 2021 and appears in the conference proceedings in 2021.

  5. 2021

    The Diffusion Schrödinger Bridge

    With Valentin De Bortoli, Jeremy Heng and Arnaud Doucet, Thornton introduces the Diffusion Schrödinger Bridge, which links diffusion models to optimal transport. The method serves as a computational optimal transport tool, the continuous state-space analogue of the Sinkhorn algorithm, and the paper is a NeurIPS 2021 spotlight.

  6. 2022

    A NeurIPS Outstanding Paper Award

    Riemannian Score-Based Generative Modelling, written with Valentin De Bortoli, Emile Mathieu, Michael Hutchinson, Yee Whye Teh and Arnaud Doucet, receives an Outstanding Paper Award at NeurIPS 2022. The paper generalizes score-based generative models from Euclidean space to Riemannian manifolds and demonstrates them on earth and climate science spherical data. In July 2022, he is first author of Riemannian Diffusion Schrödinger Bridge, which extends Riemannian score-based models beyond the first time reversal.

  7. 2023–2025

    Research scientist at Apple

    From September 2023 to February 2025, he is a research scientist in the Machine Learning Research group at Apple, after interning at NVIDIA. In April 2025, he coauthors Target Concrete Score Matching, a framework for discrete diffusion models, with Ruixiang Zhang, Shuangfei Zhai and Yizhe Zhang.

  8. Since March 2025

    Research scientist at Google DeepMind

    Thornton joins Google DeepMind as a research scientist in March 2025, in the GenMedia team, where he works on large-scale multimodal generation with flow and diffusion models. In September 2026, he is a coauthor of Simplex Diffusion Models with Arnaud Doucet and Valentin De Bortoli, all listed at Google DeepMind.

Career

  1. Since 2025Google DeepMindResearch Scientist
  2. 2023–2025AppleResearch Scientist
  3. 2016–2018BlackRockAnalyst
  4. NVIDIAIntern

Education

  1. 2023University of OxfordDPhil
  2. 2019University of OxfordProbationary research student
  3. University of Warwick

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes James Thornton. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. James Thornton | alphaXiv alphaxiv.org. Checked 22 October 2026.
  2. James Thornton jtt94.github.io. Checked 22 October 2026.
  3. Differentiable Particle Filtering via Entropy-Regularized Optimal Transport arxiv.org. Checked 22 October 2026.
  4. Differentiable Particle Filtering via Entropy-Regularized Optimal Transport proceedings.mlr.press. Checked 22 October 2026.
  5. Diffusion Schrödinger Bridge with Applications to Score-Based Generative Modeling arxiv.org. Checked 22 October 2026.
  6. Oxford Statistics Researchers win Outstanding Paper award at NeurIPS 2022 stats.ox.ac.uk. Checked 22 October 2026.
  7. Announcing the NeurIPS 2022 Awards blog.neurips.cc. Checked 22 October 2026.
  8. Riemannian Score-Based Generative Modelling proceedings.neurips.cc. Checked 22 October 2026.
  9. Riemannian Diffusion Schrödinger Bridge arxiv.org. Checked 22 October 2026.
  10. Target Concrete Score Matching: A Holistic Framework for Discrete Diffusion arxiv.org. Checked 22 October 2026.
  11. Simplex Diffusion Models arxiv.org. Checked 22 October 2026.