AI People Wiki
Portrait of Zico Kolter
Photo: Xuthoria, CC BY-SA 4.0. Credit

Builders & open source

Zico Kolter

Professor and Department Head, Machine Learning Department, Carnegie Mellon University

  • Professor and Department Head, Machine Learning Department, Carnegie Mellon University12
  • Member of the Board; chairs the Safety and Security Committee, OpenAI1
  • Member of the Board, Qualcomm1
  • Co-founder and Chief Scientist, Gray Swan AI1

Zico Kolter is a Professor and Department Head of the Machine Learning Department at Carnegie Mellon University. Kolter serves on the Board of OpenAI, where Kolter chairs the Safety and Security Committee, and on the Board of Qualcomm, and is a co-founder and Chief Scientist of Gray Swan AI, an AI security company. Kolter's group works on making deep learning more robust and safer and on understanding how data affects how models function. Kolter is a co-author of the 2023 paper "Universal and Transferable Adversarial Attacks on Aligned Language Models," which generates adversarial prompt suffixes automatically and releases its code.123

At a glance

Current roles
Professor and Department Head, Machine Learning Department +3 more
Sector
University
Education
Carnegie Mellon, Stanford
Work listed
3
Honors
4
Sources cited
9
Profile reviewed

Career

Kolter earned a PhD in computer science at Stanford University and completed a postdoctoral fellowship at MIT before joining Carnegie Mellon's School of Computer Science. Before becoming department head, Kolter was an associate professor in CMU's Computer Science Department with affiliations in Software and Societal Systems, the Robotics Institute, and Electrical and Computer Engineering, and also served in industry roles including chief scientist of AI research at Bosch.4

Kolter became head of CMU's Machine Learning Department effective June 15, 2024, as its sixth department head, succeeding Roni Rosenfeld. Kolter also sits on the boards of OpenAI, where Kolter chairs the Safety and Security Committee, and Qualcomm, co-founded the AI security company Gray Swan AI as Chief Scientist, and advises BNY.41

Research

Kolter's research spans machine learning, optimization, and control, with a focus on making deep learning safer, more robust, and more modular. Work from the group includes certified defenses against adversarial examples, such as the 2017 convex outer adversarial polytope method with Eric Wong and randomized smoothing with Jeremy Cohen and Elan Rosenfeld in 2019.4256

The group has also built optimization solvers and other modules into deep architectures, as in OptNet with Brandon Amos and Deep Equilibrium Models with Shaojie Bai and Vladlen Koltun. More recent work covers LLM security, including automatically generated adversarial suffixes that bypass model safeguards, and how training data shapes model behavior, including definitions of memorization.27834

Teaching

At CMU, Kolter teaches Graduate Artificial Intelligence (15-780) and Deep Learning Systems: Algorithms and Implementation (10-714), and has taught Practical Data Science and a course on computational methods for the smart grid. Kolter has also given tutorials on adversarial robustness (NeurIPS 2018) and implicit layers (NeurIPS 2020).9

Work and contributions

  • Universal and Transferable Adversarial Attacks on Aligned Language Models Published research

    Co-author of the 2023 paper that finds adversarial suffixes through greedy and gradient-based search, causing aligned language models to produce objectionable content, and shows that the attacks transfer to other models; the paper releases code at github.com/llm-attacks/llm-attacks.3

  • Deep Equilibrium Models Published research

    Co-author, with Shaojie Bai and Vladlen Koltun, of the 2019 paper on networks defined by the fixed point of a single layer, part of the group's work on implicit layers.89

  • Certified Adversarial Robustness via Randomized Smoothing Published research

    Co-author, with Jeremy Cohen and Elan Rosenfeld, of the 2019 paper on certifying robustness by adding noise to a classifier's inputs.6

Timeline

  1. Posted OptNet, on differentiable optimization as a layer in neural networks, with Brandon Amos.7
  2. Posted Provable defenses against adversarial examples via the convex outer adversarial polytope with Eric Wong.5
  3. Co-authored Certified Adversarial Robustness via Randomized Smoothing.6
  4. Co-authored Deep Equilibrium Models with Shaojie Bai and Vladlen Koltun.8
  5. Co-authored Universal and Transferable Adversarial Attacks on Aligned Language Models.3
  6. Became head of Carnegie Mellon University's Machine Learning Department.4

Other documented roles

Roles the sources document that are not confirmed as current.

  • Chief scientist of AI research (while on the CMU faculty), Bosch4

Honors and recognition

  • DARPA Young Faculty Award · Defense Advanced Research Projects Agency1
  • Sloan Research Fellowship · Alfred P. Sloan Foundation1
  • NeurIPS best paper award · Conference on Neural Information Processing Systems1
  • AISTATS Test of Time Award · International Conference on Artificial Intelligence and Statistics1

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

9 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Zico Kolter — BioZico Kolter · Official page · accessed Oct 21, 2026
  2. Zico Kolter — homepageZico Kolter · Official page · accessed Oct 21, 2026
  3. Universal and Transferable Adversarial Attacks on Aligned Language Models (arXiv 2307.15043)arXiv · Paper · published Jul 27, 2023 · accessed Oct 21, 2026
  4. Kolter Named Head of Carnegie Mellon University Machine Learning DepartmentCarnegie Mellon University News · Announcement · published Jun 2024 · accessed Oct 21, 2026
  5. Provable defenses against adversarial examples via the convex outer adversarial polytope (arXiv 1711.00851)arXiv · Paper · published Nov 2, 2017 · accessed Oct 21, 2026
  6. Certified Adversarial Robustness via Randomized Smoothing (arXiv 1902.02918)arXiv · Paper · published Feb 8, 2019 · accessed Oct 21, 2026
  7. OptNet: Differentiable Optimization as a Layer in Neural Networks (arXiv 1703.00443)arXiv · Paper · published Mar 1, 2017 · accessed Oct 21, 2026
  8. Deep Equilibrium Models (arXiv 1909.01377)arXiv · Paper · published Sep 3, 2019 · accessed Oct 21, 2026
  9. Zico Kolter — coursesZico Kolter · Official page · accessed Oct 21, 2026