# Durk Kingma

> Diederik Pieter "Durk" Kingma is a Dutch computer scientist whose work on the variational autoencoder and the Adam optimizer became part of the working machinery of modern machine learning. A member of OpenAI's founding team, later a research scientist at Google Brain and DeepMind, he has conducted large-scale machine-learning research at Anthropic since 2024.

- **Role:** Researcher at Anthropic
- **Organizations:** Anthropic, Google DeepMind, Google Brain, OpenAI, University of Amsterdam, Advanza, New York University
- **From:** Netherlands
- **Nationality:** Dutch
- **Education:** PhD in computer science and machine learning, cum laude; thesis: Variational Inference and Deep Learning: A New Synthesis, University of Amsterdam
- **Known for:** Co-created the variational autoencoder with Max Welling; the paper received the inaugural ICLR Test of Time Award in 2024., Co-created the Adam optimizer with Jimmy Ba; the paper received the ICLR Test of Time Award in 2025., Received the 2019 ELLIS PhD Award for research achievements during his doctorate.

## Career timeline

- **2009, 2012** — Worked as a junior research scientist in Yann LeCun's lab at New York University.
- **2010-2012** — Co-founded Advanza and served as technical lead; the company was acquired in 2016.
- **2013-2017** — Completed a cum laude PhD at the University of Amsterdam under Max Welling, with Joris Mooij as assistant supervisor; spent the summers of 2014 and 2015 collaborating at DeepMind.
- **2013** — With Max Welling, introduced Auto-Encoding Variational Bayes, the work associated with the variational autoencoder.
- **2014** — With Jimmy Ba, introduced Adam, an adaptive method for stochastic optimization.
- **2015-2018** — Joined OpenAI's founding team as a research scientist and led its algorithms team, focused on basic research.
- **2018-2024** — Worked as a research scientist at Google Brain and Google DeepMind, leading projects on generative models for text, images and video.
- **2024-present** — Conducts research on large-scale machine learning at Anthropic, working primarily from the Netherlands.
- **2026** — Presented 'Generative Models: Past, Present, Future' at the LNMB conference in the Netherlands.

## Achievements

- Co-created the variational autoencoder with Max Welling; the paper received the inaugural ICLR Test of Time Award in 2024.
- Co-created the Adam optimizer with Jimmy Ba; the paper received the ICLR Test of Time Award in 2025.
- Received the 2019 ELLIS PhD Award for research achievements during his doctorate.
- Won the 2019 Gerrit van Dijk Prize at the Dutch Data Science Awards for his dissertation.
- Earned a cum laude PhD in 2017, which his biography says was the first such distinction in the University of Amsterdam computer science department in 30 years.
- Received Google's first European Doctoral Fellowship in Deep Learning in 2015.
- Contributed to generative-model research including Glow and Variational Diffusion Models.

## Latest updates

- **2026-01** — Presented 'Generative Models: Past, Present, Future' at the 51st LNMB Conference in Soesterberg, reviewing generative-model techniques and expected near-term capabilities.
- **2025-04** — Received the ICLR 2025 Test of Time Award with Jimmy Ba for 'Adam: A Method for Stochastic Optimization.'
- **2024-10** — Joined Anthropic to conduct research on large-scale machine learning, working mostly from the Netherlands.
- **2024-05** — Received the inaugural ICLR Test of Time Award with Max Welling for 'Auto-Encoding Variational Bayes.'

## Links

- Website: https://dpkingma.com/
- LinkedIn: https://www.linkedin.com/in/durk-kingma-58b3564/
- Twitter/X: https://x.com/dpkingma
- GitHub: https://github.com/dpkingma
- YouTube: https://www.youtube.com/channel/UCBWS9Ym5IG8mDllhrnY6qqA

---

Profile page: https://yespress.io/durk-kingma
Published by YesPress — https://yespress.io
Last updated: 2026-09-26