PeopleofSuperintelligence

Researcher

James Wilson

Also known as James T. Wilson

James Wilson is a machine-learning researcher whose work on Gaussian processes and Bayesian optimization received an Outstanding Paper Honorable Mention at ICML 2020. From October 2017 to August 2022, he completed a PhD in the Sustainability and Machine Learning Group and the Department of Computing at Imperial College London, writing a thesis on decision-making with Gaussian processes under Marc Deisenroth. During his doctoral studies, he worked on the reparameterization trick for acquisition functions in December 2017 and coauthored papers on maximizing acquisition functions at NeurIPS 2018, posterior sampling at ICML 2020, and pathwise conditioning in JMLR in 2021. In 2024, he coauthored a variational framework for Gaussian process motion planning in March 2024 and published a NeurIPS 2024 paper on stopping rules for Bayesian optimization at Morgan Stanley in New York. He works in the Department of Machine Learning Research at Morgan Stanley in New York, where in 2026 he coauthored papers on the AlphaLab multi-agent research harness and tabular foundation models.

Badges

Timeline

  1. October 2017–August 2022

    A PhD on Gaussian processes

    From October 2017 to August 2022, he works toward a PhD in the Sustainability and Machine Learning Group on machine learning, Gaussian processes, Bayesian optimization and practical approximate inference. His doctoral thesis in the Department of Computing at Imperial College London, "Decision-making with Gaussian processes: sampling strategies and Monte Carlo methods", is supervised by Marc Deisenroth and introduces pathwise approaches to conditioning Gaussian processes.

  2. December 2017

    Reparameterizing acquisition functions

    In December 2017, with affiliations at Imperial College London and the University of Freiburg, he posts "The reparameterization trick for acquisition functions" on arXiv.

  3. 2018

    Maximizing acquisition functions at NeurIPS

    In 2018, he coauthors "Maximizing acquisition functions for Bayesian optimization" with Frank Hutter and Marc Deisenroth at NeurIPS. The paper shows that Monte Carlo estimates of acquisition functions are amenable to gradient-based optimization, and it identifies a family of acquisition functions, including EI and UCB, for which greedy maximization is justified.

  4. 2020

    An ICML honorable mention

    At ICML 2020, he is first author of "Efficiently sampling functions from Gaussian process posteriors", written with Viacheslav Borovitskiy, Alexander Terenin, Peter Mostowsky and Marc Deisenroth. The paper separates the prior from the data in a Gaussian process to make posterior sampling scalable, and ICML 2020 gives it an Outstanding Paper Honorable Mention.

  5. 2021

    Pathwise conditioning in JMLR

    In 2021, JMLR publishes "Pathwise Conditioning of Gaussian Processes" by James T. Wilson, Viacheslav Borovitskiy, Alexander Terenin, Peter Mostowsky and Marc Peter Deisenroth. The paper works out a pathwise view of Gaussian conditioning and tests it in settings such as global optimization and reinforcement learning.

  6. March 2024

    Gaussian process motion planning

    In March 2024, he coauthors "A Unifying Variational Framework for Gaussian Process Motion Planning" with Lucas Cosier, Rares Iordan, Sicelukwanda Zwane and Giovanni Franzese.

  7. December 2024

    Stopping Bayesian optimization

    He publishes "Stopping Bayesian Optimization with Probabilistic Regret Bounds" as sole author at NeurIPS 2024, with a December 2024 version at Morgan Stanley in New York. The paper proposes stopping rules based on the probability that a point satisfies a given set of conditions.

  8. As of October 2026

    Research at Morgan Stanley

    He works in the Department of Machine Learning Research at Morgan Stanley in New York. With Morgan Stanley colleagues, he coauthors "AlphaLab: Autonomous Multi-Agent Research Across Optimization Domains with Frontier LLMs" in March 2026 and "A Mechanistic Study of Tabular Foundation Models" with Marin Biloš, Anderson Schneider and Yuriy Nevmyvaka in May 2026. In October 2026, he coauthors "Test-Time Compute for Tabular Foundation Models: Mechanisms, Gains, and Limits" with colleagues at Morgan Stanley and the School of Computing at the University of Connecticut.

Career

  1. Since 2024Morgan Stanley

Education

  1. 2022Imperial College LondonPhD

From the biography.

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Morgan Stanley. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  2. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Imperial College London. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  3. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Imperial College London. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Imperial College London. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

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 Wilson yet. Send us a free-licensed photo.

  1. Sustainability and Machine Learning Group sml-group.cc. Checked 22 October 2026.
  2. Decision-making with gaussian processes: sampling strategies and monte carlo methods spiral.imperial.ac.uk. Checked 22 October 2026.
  3. The reparameterization trick for acquisition functions arxiv.org. Checked 22 October 2026.
  4. Maximizing acquisition functions for Bayesian optimization proceedings.neurips.cc. Checked 22 October 2026.
  5. [1805.10196] Maximizing acquisition functions for Bayesian optimization arxiv.org. Checked 22 October 2026.
  6. ICML 2020 Awards icml.cc. Checked 22 October 2026.
  7. Efficiently sampling functions from Gaussian process posteriors proceedings.mlr.press. Checked 22 October 2026.
  8. Pathwise Conditioning of Gaussian Processes jmlr.org. Checked 22 October 2026.
  9. A Unifying Variational Framework for Gaussian Process Motion Planning arxiv.org. Checked 22 October 2026.
  10. Stopping Bayesian Optimization with Probabilistic Regret Bounds proceedings.neurips.cc. Checked 22 October 2026.
  11. Stopping Bayesian Optimization with Probabilistic Regret Bounds (arXiv HTML) arxiv.org. Checked 22 October 2026.
  12. AlphaLab: Autonomous Multi-Agent Research Across Optimization Domains with Frontier LLMs arxiv.org. Checked 22 October 2026.
  13. A Mechanistic Study of Tabular Foundation Models arxiv.org. Checked 22 October 2026.
  14. Test-Time Compute for Tabular Foundation Models: Mechanisms, Gains, and Limits arxiv.org. Checked 22 October 2026.