PeopleofSuperintelligence

Researcher

Diederik P. Kingma

Also known as Diederik Kingma, Durk Kingma

Diederik P. Kingma is a computer scientist whose research spans variational inference, generative modeling, and optimization. His early work included a Utrecht University thesis on score matching, followed by a University of Amsterdam doctorate. At OpenAI, he joined the founding research team and led its Algorithms team; later, he led generative-model projects at Google Brain and DeepMind and coauthored an ICLR award-winning paper on score-based modeling. Since 2024, he has researched large-scale machine learning at Anthropic, and ICLR named his VAE and Adam papers Test of Time winners.

Now at
Anthropic
Works
49

Publication numbers from OpenAlex, as of 14 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Photo of Diederik P. Kingma

Timeline

  1. 2010–2017

    From score matching to models

    His Utrecht University master's thesis studied score matching for estimating image statistics. He completed a PhD with cum laude distinction at the University of Amsterdam in 2017, advised by Max Welling, on deep learning and generative models.

  2. 2010–2016

    A startup with an exit

    He co-founded Advanza and served as its technical lead from 2010 to 2012. The company had an exit in 2016.

  3. 2013–2014

    Reparameterizing inference

    His paper with Max Welling, Auto-Encoding Variational Bayes, introduced a stochastic variational inference method with a reparameterized lower bound for stochastic-gradient learning.

  4. 2014–2015

    Adapting stochastic updates

    With Jimmy Ba, he introduced Adam, an optimizer for stochastic objectives based on adaptive estimates of lower-order moments.

  5. 2015–2018

    Building OpenAI's research team

    At OpenAI, he joined the founding research team as a research scientist and led its Algorithms team on basic research.

  6. 2018–2024

    Generative models at Google

    At Google Brain and DeepMind, he led projects on generative models for text, images, and video. His coauthored score-based generative modeling paper received an ICLR 2021 Outstanding Paper Award.

  7. 2024–2025

    Two papers stand the test

    ICLR named Auto-Encoding Variational Bayes its Test of Time winner in 2024 and Adam its winner in 2025. Kingma coauthored the papers with Max Welling and Jimmy Ba, respectively.

  8. Since 2024

    Researching at Anthropic

    He researches large-scale machine learning at Anthropic, which he joined in 2024.

Sources and credits

Photo: dpkingma.com; Diederik P. (Durk) Kingma; creator not stated. License: Not stated by source. Source: dpkingma.com. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. News and Publications — Durk Kingma dpkingma.com. Checked 14 October 2026.
  2. Diederik P. (Durk) Kingma dpkingma.com. Checked 14 October 2026.
  3. [1312.6114] Auto-Encoding Variational Bayes arxiv.org. Checked 14 October 2026.
  4. [1412.6980] Adam: A Method for Stochastic Optimization arxiv.org. Checked 14 October 2026.
  5. Anthropic hires OpenAI co-founder Durk Kingma | TechCrunch techcrunch.com. Checked 14 October 2026.
  6. Yang Song yang-song.net. Checked 14 October 2026.
  7. ICLR 2024 Test of Time Award – ICLR Blog blog.iclr.cc. Checked 14 October 2026.
  8. Announcing the Test of Time Award Winners from ICLR 2015 – ICLR Blog blog.iclr.cc. Checked 14 October 2026.