PeopleofSuperintelligence

Researcher

Andrej Risteski

Andrej Risteski is a Carnegie Mellon machine learning researcher who studies generative models and how learning systems handle changes in their data. At Princeton, he studied computer science with doctoral advisor Sanjeev Arora; in September 2017, he joined MIT in a joint research and teaching position. With collaborators, he developed a theory of word embeddings in 2016, analyzed the limits of invariant risk minimization in 2021, and introduced a backtracking algorithm for language models in 2025. He received an NSF CAREER Award in 2023 for work on generative models and their failures in use. Since 2025, he is an associate professor in Carnegie Mellon's Machine Learning Department; in fall 2026, he teaches AI for Scientific Computing.

Works
89
Citations
1,216
h-index
15

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

  1. 2008–2017

    Computer science at Princeton

    He earns a bachelor's degree in computer science at Princeton University in 2012. His doctoral studies in Princeton's Computer Science Department run from 2012 to 2017, with Sanjeev Arora as his advisor.

  2. 2016

    A theory of word embeddings

    In 2016, he coauthors “A Latent Variable Model Approach to PMI-based Word Embeddings” with Sanjeev Arora, Yuanzhi Li, Yingyu Liang and Tengyu Ma. The paper introduces a generative model that explains mathematical structure in word embeddings and provides a theoretical basis for word2vec and GloVe.

  3. September 2017

    A joint position at MIT

    In September 2017, he joins MIT's Institute for Data, Systems, and Society as a Norbert Wiener Fellow and its Applied Mathematics department as an instructor.

  4. 2019

    Joining Carnegie Mellon's faculty

    He joins Carnegie Mellon University's Machine Learning Department as an assistant professor in 2019.

  5. 2021

    The limits of invariant learning

    At ICLR 2021, he coauthors “The Risks of Invariant Risk Minimization” with Elan Rosenfeld and Pradeep Ravikumar. The paper analyzes invariant risk minimization and shows that the method can fail when test data differ from the training distribution.

  6. 2023

    An NSF CAREER Award

    He receives an NSF CAREER Award in 2023, with nearly $530,000 to study the mathematical foundations of generative models and their failures when put to use.

  7. October 2025

    Backtracking for language models

    In October 2025, he coauthors “Taming Imperfect Process Verifiers: A Sampling Perspective on Backtracking”. Its VGB sampling algorithm uses probabilistic backtracking to make text generation more robust to errors in tools that check partial generations.

  8. Since 2025

    Associate professor at Carnegie Mellon

    In 2025, he becomes an associate professor in Carnegie Mellon University's Machine Learning Department. His research covers generative models, representation learning and out-of-distribution generalization. In fall 2026, he teaches 10-749, AI for Scientific Computing.

Career

  1. Since 2025Carnegie Mellon UniversityAssociate Professor
  2. 2017–2019Massachusetts Institute of TechnologyInstructor

Education

  1. 2017Princeton UniversityPhD
  2. 2012Princeton UniversityBachelor's degree in computer science

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

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 Andrej Risteski yet. Send us a free-licensed photo.

  1. Andrej Risteski - Machine Learning Research andrejristeski.org. Checked 27 October 2026.
  2. A Latent Variable Model Approach to PMI-based Word Embeddings - ACL Anthology aclanthology.org. Checked 27 October 2026.
  3. New provable techniques for learning and inference in probabilistic graphical models – MIT Statistics and Data Science Center stat.mit.edu. Checked 27 October 2026.
  4. Recent Faculty Hires / 2019 - SCS Business Office - SCS Computing Facilities scsbusinessoffice.cs.cmu.edu. Checked 27 October 2026.
  5. [2010.05761] The Risks of Invariant Risk Minimization arxiv.org. Checked 27 October 2026.
  6. SCS Faculty Receive More Than $4.5M in NSF CAREER Awards cs.cmu.edu. Checked 27 October 2026.
  7. [2510.03149] Taming Imperfect Process Verifiers: A Sampling Perspective on Backtracking arxiv.org. Checked 27 October 2026.
  8. Kempner Institute Announces 2026-27 Kempner Seminar Series Lineup - Kempner Institute kempnerinstitute.harvard.edu. Checked 27 October 2026.
  9. 10-749 Fall 2026: AI for Scientific Computing andrejristeski.github.io. Checked 27 October 2026.