PeopleofSuperintelligence

Lab leader

Radu Soricut

Also known as R. Soricut

Radu Soricut is a computer scientist whose work spans natural language generation, machine translation, and multimodal AI. He earned a PhD at USC, where he worked with Daniel Marcu and Kevin Knight, and coauthored a 2006 WIDL-expressions paper with Marcu on machine translation and summarization. In 2012, he was a manager of application science and engineering and a senior research scientist at SDL International; he also coauthored a paper on Language Weaver quality-estimation systems. His work included an unsupervised morphology method in 2015, a co-lead role for Gemini multimodal vision in 2023, and coauthorship of the Gemini Robotics report in 2025. As of October 2026, he is a Distinguished Scientist and Senior Research Director at Google DeepMind, where he leads teams focused on language modeling, machine learning, and multimodal models.

Timeline

  1. 2006

    A PhD in language generation

    He completes a PhD at USC, where he works with Daniel Marcu and Kevin Knight. In 2006, he and Marcu publish a WIDL-expressions paper on machine translation and summarization.

  2. 2012

    Quality estimation at SDL

    At META-FORUM 2012, he is manager of application science and engineering and senior research scientist at SDL International. In 2012, he coauthors the SDL Language Weaver quality-estimation paper with Nguyen Bach and Ziyuan Wang.

  3. 2015

    Inducing word morphology

    In 2015, he and Franz Och present an unsupervised method for inducing morphological transformations between words. The method relies on regularities in high-dimensional vector spaces.

  4. 2023

    Co-leading Gemini vision

    In 2023, he co-leads multimodal vision for Gemini. Gemini is a model family for image, audio, video, and text understanding.

  5. 2025

    Gemini Robotics models

    In 2025, he coauthors the Gemini Robotics report, which introduces AI models built on Gemini 2.0 for robotics. The report presents Gemini Robotics as a vision-language-action model for direct robot control.

  6. As of October 2026

    Google DeepMind research director

    As of October 2026, he is a Distinguished Scientist and Senior Research Director at Google DeepMind. He leads teams that conduct natural language modeling and machine-learning research and build large models for multimodal language and vision understanding and generation.

Sources and credits

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

  1. Radu Soricut research.google. Checked 15 October 2026.
  2. Alumni - Artificial Intelligence isi.edu. Checked 15 October 2026.
  3. Stochastic Language Generation Using WIDL-Expressions and its Application in Machine Translation and Summarization aclanthology.org. Checked 15 October 2026.
  4. The SDL Language Weaver Systems in the WMT12 Quality Estimation Shared Task aclanthology.org. Checked 15 October 2026.
  5. META-FORUM 2012 - Report mt-archive.net. Checked 15 October 2026.
  6. Unsupervised Morphology Induction Using Word Embeddings research.google. Checked 15 October 2026.
  7. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 15 October 2026.
  8. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 15 October 2026.
  9. Gemini Robotics: Bringing AI into the Physical World arxiv.org. Checked 15 October 2026.