Durk Kingma

Diederik P. Kingma
e-mail: dpkingma [at] gmail [dot] com
Brief Bio | Publications | PhD Thesis | Demos | Links | Honors
My portrait

Science, and intelligent behaviour in general, involves learning from data: finding predictive models (theories) that are consistent with observations. Machine learning is the study of algorithms for automating this learning process with computer systems. Due to the generality of the problem of learning from data, it is difficult to overstate how impactful machine learning is: its algorithms and tools are used throughout the sciences and industry. And due to the exponential growth of computational resources, its potential keeps expanding.

I'm a machine learning researcher, since 2018 at Google. My contributions include the Variational Autoencoder (VAE), the Adam optimizer, Inverse Autoregressive Flow (IAF), and Glow. More generally my main research interests are at the intersection of deep learning with topics such as generative models, variational (Bayesian) inference, stochastic optimization, and identifiability. I obtained a PhD (cum laude) from University of Amsterdam in 2017, and was part of the founding team of OpenAI in 2015. Before that, I co-founded Advanza which got acquired in 2016.

My formal first name is Diederik, but people also call me Durk, which is a Frisian name that is pronounced like the English name Dirk.

I live in San Francisco.

Brief Bio


See my Google Scholar profile for a complete list.

Variational Inference and Deep Learning: A New Synthesis

Ph.D. Thesis. Download at Dropbox or at UvA.


Some research demos that I (co-)developed:
These places are updated more frequently than this website:

Awards and Honors

The end. Want more? Click one of the blue links.