{"version":1,"id":"aleksander-madry","title":"Aleksander Madry","summary":"Aleksander Mądry is a computer scientist and MIT professor known for research in machine learning, adversarial robustness, optimization, and graph algorithms.","content":"**Aleksander Mądry** is a computer scientist whose research spans machine learning, optimization, graph algorithms, and theoretical computer science. He is a professor in the Massachusetts Institute of Technology’s (MIT) Department of Electrical Engineering and Computer Science (EECS), a member of its Computer Science and Artificial Intelligence Laboratory (CSAIL), director of the MIT Center for Deployable Machine Learning, and a faculty co-lead of the MIT AI Policy Forum.[\\[1\\]](#cite-id-8bvgdsuuaw)​[\\[2\\]](#cite-id-rxaks3kw7c)​\n\n## Education and Career\n\nMądry studied computer science at MIT from 2007 to 2011 and earned his Ph.D. there in 2011. His doctoral thesis, “From Graphs to Matrices, and Back: New Techniques for Graph Algorithms,” received an ACM Doctoral Dissertation Award Honorable Mention and MIT’s George M. Sprowls Award for best doctoral theses in computer science.[\\[1\\]](#cite-id-8bvgdsuuaw)​[\\[2\\]](#cite-id-rxaks3kw7c)​\n\nFrom July 2011 to June 2012, he was a postdoctoral researcher at Microsoft Research New England. He then joined the faculty at EPFL, where he was an assistant professor from July 2012 to January 2015. He became a professor at MIT in February 2015.[\\[2\\]](#cite-id-rxaks3kw7c) Mądry’s research interests include developing machine-learning methods that are reliable and well understood, and that can be deployed safely and responsibly to support real-world decision-making.[\\[1\\] ](#cite-id-8bvgdsuuaw)​\n\n## Research\n\n### Graph Algorithms and Optimization\n\nMądry’s early work focused on graph algorithms, flows, matchings, routing, and online algorithms. His research used electrical-flow methods in problems involving graph connectivity and network flows. In “_Electrical Flows, Laplacian Systems, and Faster Approximation of Maximum Flow in Undirected Graphs_,” coauthored with Paul Christiano, Jonathan Kelner, Daniel Spielman, and Shang-Hua Teng, he connected electrical flows and Laplacian systems with faster approximation of maximum flow in undirected graphs. The paper received the Best Paper Award at the 2011 ACM Symposium on Theory of Computing (STOC) and was invited to the _Journal of the ACM_.[\\[1\\]](#cite-id-8bvgdsuuaw)​[\\[3\\]](#cite-id-odonstar0m)​\n\nOther work from this period included “_Faster Generation of Random Spanning Trees_,” “_Fast Approximation Algorithms for Cut-based Problems in Undirected Graphs_,” and “Fast_er Approximation Schemes for Fractional Multicommodity Flow Problems via Dynamic Graph Algorithms_.” His 2010 paper on an approximation algorithm for the asymmetric traveling salesman problem, coauthored with Arash Asadpour, Michel Goemans, Shayan Oveis Gharan, and Amin Saberi, received the Best Paper Award at the Symposium on Discrete Algorithms (SODA).[\\[1\\]](#cite-id-8bvgdsuuaw)​\n\n### Machine-Learning Robustness and Evaluation\n\nFrom 2017 onward, Mądry’s publications increasingly addressed deep learning, including how models respond to adversarial inputs and how robustness relates to accuracy and generalization. “_Towards Deep Learning Models Resistant to Adversarial Attacks_,” published at ICLR in 2018, was followed by “_Adversarially Robust Generalization Requires More Data_” and “_Robustness May Be at Odds with Accuracy_.” These papers examined adversarial robustness in relation to model training, data requirements, and accuracy. “_Adversarial Examples Are Not Bugs, They Are Features_,” published at NeurIPS in 2019, was presented as a spotlight paper. Other work from the period examined black-box attacks using bandit methods and priors, spatial robustness, and the use of a robust classifier for image synthesis. Mądry also coauthored “_On Evaluating Adversarial Robustness_,” a 2019 study of methods for assessing model robustness.[\\[1\\]](#cite-id-8bvgdsuuaw)​[\\[3\\]](#cite-id-odonstar0m)​\n\n## Work at OpenAI\n\nIn a January 2024 LinkedIn post, Mądry described OpenAI’s early work, as part of its Preparedness effort, assessing risks associated with large language models enabling biological threat creation. He said the work included an initial blueprint for evaluating that risk.[\\[2\\]](#cite-id-rxaks3kw7c)​\n\nMądry is also listed as a coauthor of the 2024 GPT-4o system card and the 2024 OpenAI o1 system card.[\\[3\\]](#cite-id-odonstar0m)​\n\n## Teaching and Academic Leadership\n\nAt MIT, Mądry has taught courses in algorithms, theoretical computer science, optimization, machine learning, and decision-making. His courses include Design and Analysis of Algorithms, Advanced Algorithms, Science of Deep Learning, Online Decision Making: Optimization, Control and Games, and AI, Decision Making, and Society. He taught Design and Analysis of Algorithms across multiple terms from Spring 2017 through Spring 2023.[\\[1\\]](#cite-id-8bvgdsuuaw)​\n\nHis academic leadership includes directing the MIT Center for Deployable Machine Learning and serving as a faculty co-lead of the MIT AI Policy Forum.[\\[1\\]](#cite-id-8bvgdsuuaw)​\n","categories":[{"id":"people","title":"people"}],"tags":[{"id":"Researcher"}],"images":[{"id":"QmXskfAzEqj1W9iQmoWAL2K5AH39iAewwhFdeym44xU6AZ","type":"image/jpeg, image/png"}],"media":[],"metadata":[{"id":"references","value":"[{\"id\":\"8bvgdsuuaw\",\"url\":\"https://madry.mit.edu/\",\"description\":\"MIT faculty profile\",\"timestamp\":1792823254382},{\"id\":\"rxaks3kw7c\",\"url\":\"https://www.linkedin.com/in/aleksander-madry-61115b233\",\"description\":\"Professional profile\",\"timestamp\":1792823254382},{\"id\":\"odonstar0m\",\"url\":\"https://scholar.google.com/citations?user=SupjsEUAAAAJ&hl=en\",\"description\":\"Google Scholar profile\",\"timestamp\":1792823254382}]"},{"id":"twitter_profile","value":"https://twitter.com/aleks_madry"},{"id":"linkedin_profile","value":"https://www.linkedin.com/in/aleksander-madry-61115b233/"},{"id":"main-image-origin","value":"{\"imageId\":\"QmXskfAzEqj1W9iQmoWAL2K5AH39iAewwhFdeym44xU6AZ\",\"originalId\":\"QmXskfAzEqj1W9iQmoWAL2K5AH39iAewwhFdeym44xU6AZ\"}"},{"id":"commit-message","value":"Add Aleksander Madry wiki"}],"events":[],"linkedWikis":{"founders":[],"blockchains":[],"speakers":[]},"user":{"id":"0x8af7a19a26d8fbc48defb35aefb15ec8c407f889"},"author":{"id":"0x4a2bc97afbf41a2cbf4b4628f63420682ff4cf2a"},"language":"en","operator":{"id":"0x4a2bc97afbf41a2cbf4b4628f63420682ff4cf2a"}}