PeopleofSuperintelligence

Researcher

Igor Mordatch

Igor Mordatch is a researcher whose work connects robotics, reinforcement learning and multi-agent systems. At the University of Washington, he earned a PhD in computer science and engineering in 2015 and worked as a postdoctoral fellow at UC Berkeley. A 2017 study he coauthored examined how agents develop grounded language, and his continuous meta-learning paper received the ICLR 2018 Best Paper Award. He also coauthored Neural MMO and later contributed to Melting Pot, RT-2 and research on self-improving embodied foundation models. In 2026, he coauthored TokenSwap to study and reduce modality gaps in multimodal language models; as of October 2026, he is a research scientist at Google DeepMind and a lecturer at UC Berkeley.

Timeline

  1. 2015

    A robotics doctorate

    In 2015, he earns a PhD in computer science and engineering from the University of Washington and works as a postdoctoral fellow at UC Berkeley. His UW CSE work builds on collaboration with PhD advisor Emo Todorov, who leads the Movement Control Laboratory, and Zoran Popović, director of the Center for Game Science; a simulated robot model lets an algorithm search for efficient movements from an end goal.

  2. 2017

    Language among agents

    In 2017, he coauthors a study on how grounded compositional language can emerge in multi-agent populations. Its learning environment brings out a basic vocabulary and syntax.

  3. 2018

    An ICLR Best Paper

    At ICLR 2018, his paper on continuous adaptation via meta-learning wins a Best Paper Award.

  4. 2019

    Neural MMO

    In 2019, he coauthors Neural MMO, a persistent environment where a large, variable population of agents learns robust combat and navigation policies.

  5. 2021

    Testing social generalisation

    In 2021, he coauthors Melting Pot, an evaluation suite with 21 multi-agent games and more than 85 test scenarios to measure how agents handle unfamiliar social situations.

  6. July 2023

    Robot control with RT-2

    In July 2023, he contributes to RT-2, a vision-language-action model that shows improved generalization and semantic and visual understanding beyond the robot data it was exposed to.

  7. 2025

    Self-improving robot models

    In 2025, he coauthors Self-Improving Embodied Foundation Models, which uses steps-to-go predictions to produce rewards and a success detector so robots can practice downstream tasks with minimal human supervision.

  8. As of October 2026

    Research at DeepMind, teaching at Berkeley

    In May 2026, he coauthors TokenSwap, which measures the modality gap between semantically equivalent text-only and image-interleaved inputs and tests a method to reduce it. As of October 2026, he is a research scientist at Google DeepMind and a lecturer at UC Berkeley. His career also includes a research scientist role at OpenAI.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Sources and credits

Photo: We have no published photo of Igor Mordatch yet. Send us a free-licensed photo.

  1. Allen School News Archive » UW CSE Ph.D. alum Igor Mordatch aims to teach robots to learn for themselves news.cs.washington.edu. Checked 19 October 2026.
  2. [1703.04908] Emergence of Grounded Compositional Language in Multi-Agent Populations arxiv.org. Checked 19 October 2026.
  3. Publications people.eecs.berkeley.edu. Checked 19 October 2026.
  4. [1903.00784] Neural MMO: A Massively Multiagent Game Environment for Training and Evaluating Intelligent Agents arxiv.org. Checked 19 October 2026.
  5. Melting Pot: an evaluation suite for multi-agent reinforcement learning — Google DeepMind deepmind.google. Checked 19 October 2026.
  6. RT-2: New model translates vision and language into action — Google DeepMind deepmind.google. Checked 19 October 2026.
  7. [2509.15155] Self-Improving Embodied Foundation Models arxiv.org. Checked 19 October 2026.
  8. [2607.28640] TokenSwap: Benchmarking and Reducing the Modality Gap in Multimodal LLMs arxiv.org. Checked 19 October 2026.
  9. Igor Mordatch simons.berkeley.edu. Checked 19 October 2026.