Barnabás Póczos

Also known as Barnabas Poczos

Barnabás Póczos is a machine-learning researcher whose work connects statistics and machine learning. He earned degrees in applied mathematics and computer science at Eötvös Loránd University, then held postdoctoral roles at the University of Alberta and Carnegie Mellon. His coauthored work includes Deep Sets, a neural-network architecture for data represented as sets, and a 2019 paper on density estimation for GANs that received a NeurIPS Outstanding Paper honorable mention. By 2025, he was a co-founder of Ansatz AI, whose Hierarchical Machine Learning software was released with Procter & Gamble as a development partner; as of October 2026, he is an associate professor and Ph.D. program co-director at Carnegie Mellon.

Now at: Carnegie Mellon University
Badges: Founder

Timeline

2001–2007

Degrees in Budapest

He earns an M.Sc. in applied mathematics in 2001 and a Ph.D. in computer science in 2007 at Eötvös Loránd University in Budapest.
2007–2012

Postdoctoral research

From 2007 to 2010, he works as a postdoctoral fellow in the RLAI group at the University of Alberta. In 2010, he moves to Carnegie Mellon's Auton Lab, where he holds a postdoctoral fellowship through 2012.
2012

A Yahoo! ACE Award

He receives Yahoo!'s Academic Career Enhancement (ACE) Award in 2012.
2017

Deep Sets

In 2017, he coauthors Deep Sets, which characterizes permutation-invariant objectives and presents a deep-network architecture for machine-learning tasks defined on sets.
2019

Implicit kernel learning

In 2019, he coauthors Implicit Kernel Learning, whose applications include MMD GANs and supervised learning.
2019

GAN convergence rates

With Ananya Uppal and Shashank Singh, he studies nonparametric density estimation under Besov IPM losses and bounds GAN statistical error. The paper receives an Honorable Mention for NeurIPS 2019's Outstanding Paper Award.
By 2025

Co-founding Ansatz AI

By 2025, he is a co-founder of the Carnegie Mellon startup Ansatz AI, which releases Hierarchical Machine Learning software for formulated-product design with Procter & Gamble as a development partner.
As of October 2026

Associate professor at CMU

As of October 2026, he is an associate professor and co-director of the Ph.D. program in Carnegie Mellon's Machine Learning Department. He studies statistical questions that shape machine-learning methods and teaches Deep Learning and Generative AI Theory in fall 2026.