AI People Wiki

Builders & open source

Kristina Toutanova

Research scientist, Google DeepMind

Kristina Toutanova is a research scientist at Google, listed with Google DeepMind on an ICLR 2026 paper, and an affiliate faculty member at the University of Washington, where the Department of Linguistics lists Toutanova as an Affiliate Assistant Professor. Before joining Google in 2017, Toutanova was a researcher at Microsoft Research. Google Research's profile describes work on modeling the structure of natural language with machine learning, most recently in representation learning, question answering, information retrieval, semantic parsing, and knowledge base completion, and notes that Toutanova is a past co-editor in chief of TACL. Toutanova is a co-author of BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, whose code and pretrained models are released under the Apache 2.0 license in the google-research/bert repository.12345

At a glance

Current roles
Research scientist +1 more
Sector
Company
Education
Stanford, UW
Work listed
1
Sources cited
9
Profile reviewed

Career

Kristina Toutanova completed a PhD in Computer Science at Stanford University with Christopher Manning, finishing in 2005, and then worked as a researcher at Microsoft Research. Toutanova joined Google in 2017 as a research scientist in Google AI Language and is an affiliate faculty member at the University of Washington, where the Department of Linguistics lists Toutanova as an Affiliate Assistant Professor.13

Toutanova is a past co-editor in chief of the journal Transactions of the Association for Computational Linguistics (TACL) and was a program co-chair of ACL 2014.1

Research

Toutanova's research models the structure of natural language with machine learning, most recently in representation learning, question answering, information retrieval, semantic parsing and knowledge base completion. Toutanova co-authored BERT (2018) with Jacob Devlin, Ming-Wei Chang and Kenton Lee, and the 2019 paper on latent retrieval for weakly supervised open-domain question answering.146

Later work includes a 2020 study of compositional generalization in semantic parsing, Pix2Struct (2022), which pretrains visual language models by parsing screenshots, and QUEST (2023), a retrieval dataset of queries with implicit set operations. A 2025 paper with Peter Shaw, James Cohan and Jacob Eisenstein studies description-length objectives for transformers grounded in Kolmogorov complexity.7892

Work and contributions

  • BERT Open-source code

    Co-author with Jacob Devlin, Ming-Wei Chang, and Kenton Lee of the 2018 BERT paper; the google-research/bert repository, which links the paper and lists Toutanova in its citation, releases TensorFlow code and pretrained models under the Apache 2.0 license.45

Timeline

  1. Completed a PhD in Computer Science at Stanford University with Christopher Manning.31
  2. Served as a program co-chair of ACL 2014.1
  3. Joined Google as a research scientist after working at Microsoft Research.1
  4. Co-authored BERT, on pre-training deep bidirectional transformers for language understanding.4
  5. Co-authored "Latent Retrieval for Weakly Supervised Open Domain Question Answering".6
  6. Co-authored a study of compositional generalization and natural language variation in semantic parsing.7
  7. Co-authored Pix2Struct, which pretrains visual language understanding by parsing screenshots.8
  8. Co-authored QUEST, a retrieval dataset of entity-seeking queries with implicit set operations.9
  9. Co-authored a paper on asymptotically optimal description length objectives for transformers.2

Other documented roles

Roles the sources document that are not confirmed as current.

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. Kristina N Toutanova — Google Research people profileGoogle Research · Official page · accessed Oct 21, 2026
  2. Bridging Kolmogorov Complexity and Deep Learning: Asymptotically Optimal Description Length Objectives for Transformers (arXiv 2509.22445, ICLR 2026)arXiv · Paper · published Sep 26, 2025 · accessed Oct 21, 2026
  3. Kristina Toutanova — University of Washington Department of Linguistics people pageUniversity of Washington Department of Linguistics · Official page · accessed Oct 21, 2026
  4. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (arXiv 1810.04805)arXiv · Paper · published Oct 11, 2018 · accessed Oct 21, 2026
  5. google-research/bert README (Apache 2.0)Google Research (GitHub repository) · Repository · accessed Oct 21, 2026
  6. Latent Retrieval for Weakly Supervised Open Domain Question Answering (arXiv 1906.00300)arXiv · Paper · published Jun 1, 2019 · accessed Oct 21, 2026
  7. Compositional Generalization and Natural Language Variation: Can a Semantic Parsing Approach Handle Both? (arXiv 2010.12725)arXiv · Paper · published Oct 24, 2020 · accessed Oct 21, 2026
  8. Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding (arXiv 2210.03347)arXiv · Paper · published Oct 7, 2022 · accessed Oct 21, 2026
  9. QUEST: A Retrieval Dataset of Entity-Seeking Queries with Implicit Set Operations (arXiv 2305.11694)arXiv · Paper · published May 19, 2023 · accessed Oct 21, 2026