PeopleofSuperintelligence

Researcher

Michalis Titsias

Also known as Michalis K. Titsias

Michalis Titsias is a machine learning researcher at Google DeepMind whose work covers variational inference, sparse Gaussian processes, Monte Carlo sampling and gradient estimation. After studying computer science at the University of Ioannina, they earned a PhD at the University of Edinburgh in 2005 under Christopher Williams, and they held postdoctoral posts at the University of Oxford and the University of Manchester and an assistant professorship at the Athens University of Economics and Business. Early papers include a 2009 variational method for sparse Gaussian processes and the 2014 Doubly Stochastic Variational Bayes algorithm, written with Miguel Lázaro-Gredilla. In 2022, they coauthored Gradient Estimation with Discrete Stein Operators, a NeurIPS 2022 Outstanding Paper, and in 2024 they coauthored Simplified and Generalized Masked Diffusion for Discrete Data. They are a research scientist at Google DeepMind in London.

Badges

Timeline

  1. 2005

    A PhD at Edinburgh

    After studying computer science at the University of Ioannina in Greece, they complete a PhD at the University of Edinburgh. The thesis, Unsupervised Learning of Multiple Objects in Images, is advised by Christopher Williams at the Institute for Adaptive and Neural Computation and appears in June 2005.

  2. 2009

    Variational sparse Gaussian processes

    In 2009, they publish Variational Learning of Inducing Variables in Sparse Gaussian Processes at the Twelfth International Conference on Artificial Intelligence and Statistics. The method treats the inducing inputs as variational parameters and selects them, together with the kernel hyperparameters, by maximizing a lower bound of the log marginal likelihood.

  3. 2014

    Doubly stochastic variational Bayes

    In 2014, the Proceedings of the 31st International Conference on Machine Learning carry Doubly Stochastic Variational Bayes for non-Conjugate Inference, which they write with Miguel Lázaro-Gredilla. The paper proposes a variational inference algorithm based on stochastic optimisation for Bayesian non-conjugate inference in continuous parameter spaces.

  4. By 2019

    From AUEB to DeepMind

    They work as a postdoctoral researcher at the University of Oxford and the University of Manchester and as an assistant professor in the Department of Informatics at the Athens University of Economics and Business. By 2019, they are a full-time research scientist at DeepMind in London, and in December 2019 they speak on Gradient-based Adaptive Markov Chain Monte Carlo at the Department of Engineering in Cambridge.

  5. 2022

    A NeurIPS Outstanding Paper

    In 2022, they coauthor Gradient Estimation with Discrete Stein Operators with Jiaxin Shi, Yuhao Zhou, Jessica Hwang and Lester Mackey, which receives a NeurIPS 2022 Outstanding Paper Award. The paper introduces a variance reduction technique based on Stein operators for discrete distributions and uses it to build control variates for the REINFORCE leave-one-out estimator.

  6. 2024

    Masked diffusion for discrete data

    In 2024, they coauthor Simplified and Generalized Masked Diffusion for Discrete Data at NeurIPS with Jiaxin Shi, Kehang Han, Zhe Wang and Arnaud Doucet. The paper shows that the continuous-time variational objective of masked diffusion models is a weighted integral of cross-entropy losses, and the official JAX implementation, MD4, is published in the google-deepmind GitHub organization.

  7. 2025

    Tighter sparse Gaussian process bounds

    In 2025, they publish New Bounds for Sparse Variational Gaussian Processes in the Proceedings of the 42nd International Conference on Machine Learning. In Gaussian process regression, the paper derives a tractable collapsed bound that is tighter than the previous bound.

  8. As of October 2026

    Research scientist at DeepMind

    They are a full-time research scientist at Google DeepMind in London, with research interests in machine learning and statistical methodology. Their recent work includes Learning-Order Autoregressive Models with Application to Molecular Graph Generation, an ICML 2025 paper.

Career

  1. NowGoogle DeepMindResearch Scientist
  2. University of OxfordPostdoctoral Researcher
  3. University of ManchesterPostdoctoral Researcher
  4. Athens University of Economics and BusinessAssistant Professor

Education

  1. 2005University of EdinburghPhD
  2. University of Ioannina

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Michalis Titsias. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Michalis Titsias. 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.
  3. 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.
  4. 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.

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 Michalis Titsias yet. Send us a free-licensed photo.

  1. Michalis Titsias mtitsias.github.io. Checked 21 October 2026.
  2. Unsupervised Learning of Multiple Objects in Images era.ed.ac.uk. Checked 21 October 2026.
  3. Variational Learning of Inducing Variables in Sparse Gaussian Processes proceedings.mlr.press. Checked 21 October 2026.
  4. Doubly Stochastic Variational Bayes for non-Conjugate Inference proceedings.mlr.press. Checked 21 October 2026.
  5. [AI@AUEB/ML Group Meeting - Seminar Series] Invited talk: Simplified and Generalized Masked Diffusion for Discrete Data archimedesai.gr. Checked 21 October 2026.
  6. Michalis K. Titsias, Athens University of Economics and Business www2.aueb.gr. Checked 21 October 2026.
  7. talks.cam : LR11, Department of Engineering talks.cam.ac.uk. Checked 21 October 2026.
  8. Recent publications · Michalis Titsias mtitsias.github.io. Checked 21 October 2026.
  9. Research Focus: Week of December 5, 2022 - Microsoft Research microsoft.com. Checked 21 October 2026.
  10. Simplified and Generalized Masked Diffusion for Discrete Data proceedings.neurips.cc. Checked 21 October 2026.
  11. GitHub - google-deepmind/md4: Official Jax Implementation of MD4 Masked Diffusion Models github.com. Checked 21 October 2026.
  12. New Bounds for Sparse Variational Gaussian Processes proceedings.mlr.press. Checked 21 October 2026.