PeopleofSuperintelligence

Researcher

Nicolas Flammarion

Nicolas Flammarion is a machine learning researcher at EPFL whose work on optimization and the training of deep networks reaches into the safety of large language models. He earned his PhD at the École Normale Supérieure in Paris in 2017, received the Fondation Mathématique Jacques Hadamard prize for the best PhD thesis in optimization in 2018, and was a postdoctoral fellow at UC Berkeley before becoming a tenure-track assistant professor at EPFL in 2019. In 2021, he shared a NeurIPS Outstanding Paper Award for the continuized acceleration of gradient descent. In 2024, he coauthored a paper on jailbreaking safety-aligned language models and received a Trust & Safety Google Research Award. He is an associate professor of computer science at EPFL, with research on learning problems at the intersection of machine learning, statistics and optimization.

Now at
EPFL

Timeline

  1. 2017

    PhD at École Normale Supérieure

    He receives his PhD in 2017 from the École Normale Supérieure in Paris, where Alexandre d'Aspremont and Francis Bach advise him.

  2. 2018

    Hadamard prize for best thesis

    In 2018, he receives the prize of the Fondation Mathématique Jacques Hadamard for the best PhD thesis in the field of optimization.

  3. By 2019

    Postdoc at UC Berkeley

    By 2019, he is a postdoctoral fellow in computer science at UC Berkeley, hosted by Michael I. Jordan. In March 2019, he gives an IC Colloquium at EPFL as a faculty candidate, showing how the underdamped Langevin algorithm performs accelerated gradient descent on the KL divergence.

  4. 2019

    Tenure-track professor at EPFL

    In July 2019, the Board of the Swiss Federal Institutes of Technology names him Tenure Track Assistant Professor of Computer Science and Communication Systems in EPFL's School of Computer and Communication Sciences, for research on the design, analysis and optimisation of sampling procedures for machine learning. He joins EPFL in 2019 and heads its Theory of Machine Learning Laboratory (TML).

  5. 2021

    Continuized acceleration wins at NeurIPS

    In 2021, the paper "Continuized Accelerations of Deterministic and Stochastic Gradient Descents, and of Gossip Algorithms", which he writes with Mathieu Even, Raphaël Berthier, Francis Bach and others, wins a NeurIPS Outstanding Paper Award, one of six chosen from almost 9,000 submissions. It describes a "continuized" version of Nesterov's accelerated gradient method whose gradient updates occur at random times set by a Poisson point process.

  6. October 2023

    A paper on weight decay

    In October 2023, he coauthors "Why Do We Need Weight Decay in Modern Deep Learning?" with Francesco D'Angelo, Maksym Andriushchenko and Aditya Varre. The paper shows that weight decay strengthens the implicit regularization of SGD on vision tasks trained with multipass SGD, and that it balances the bias-variance tradeoff for large language models trained for nearly one epoch.

  7. 2024

    Jailbreak attacks and a Google award

    In 2024, he coauthors "Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks" with Maksym Andriushchenko and Francesco Croce, which appears at ICLR 2025. Their random search on a prompt suffix reaches a 100% attack success rate, judged by GPT-4, on models including Llama-3-Instruct-8B, GPT-3.5 and GPT-4o. He also receives a Trust & Safety Google Research Award in 2024.

  8. As of October 2026

    Associate professor at EPFL

    In May 2026, EPFL announces his promotion from tenure-track assistant professor to associate professor of computer science and communication systems in the School of Computer and Communication Sciences. His research covers learning problems at the intersection of machine learning, statistics and optimization.

Career

  1. Since 2019EPFLTenure Track Assistant Professor of Computer Science and Communication Systems, Associate Professor of Computer Science and Communication Systems
  2. Until 2019University of California, BerkeleyPostdoctoral Fellow in Computer Science

Education

  1. 2017École Normale SupérieurePhD

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Nicolas Flammarion. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  2. Google has the highest combined h-index among AI organizationsIncludes EPFL. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. The median h-index is 48 at universities and 14 at frontier AI labsIncludes EPFL. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. The United States employs an estimated 48% of people shaping AIIncludes EPFL. We estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.

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 Nicolas Flammarion yet. Send us a free-licensed photo.

  1. EPFL - Nicolas Flammarion people.epfl.ch. Checked 22 October 2026.
  2. IC Colloquium: Optimization and Monte Carlo Sampling in Large Scale Machine Learning - EPFL memento.epfl.ch. Checked 22 October 2026.
  3. Two new IC professors appointed - EPFL actu.epfl.ch. Checked 22 October 2026.
  4. “Trying to understand the world really drives me” - EPFL actu.epfl.ch. Checked 22 October 2026.
  5. Announcing the NeurIPS 2021 Award Recipients blog.neurips.cc. Checked 22 October 2026.
  6. [2310.04415] Why Do We Need Weight Decay in Modern Deep Learning? arxiv.org. Checked 22 October 2026.
  7. [2404.02151] Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks arxiv.org. Checked 22 October 2026.
  8. Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks | ML Anthology mlanthology.org. Checked 22 October 2026.
  9. Appointment of EPFL professors - EPFL actu.epfl.ch. Checked 22 October 2026.