PeopleofSuperintelligence

Lab leader

Matthew Henderson

Also known as Matt Henderson, matthen

Matthew Henderson is a spoken-dialogue researcher and PolyAI research leader whose work spans dialogue tracking, conversational models and voice agents. He studied mathematics at Cambridge and speech and language processing at Edinburgh, pursued dialogue-systems research under Steve Young, and was selected for a Global Google PhD Fellowship in 2013. At Google he coauthored Smart Reply research; at Reka he coauthored a multimodal-model report; at PolyAI he coauthored work on conversational datasets and ConveRT, then introduced Raven v2 and Dialog-RSN-1. He is VP of Research at PolyAI and makes mathematical animations, including a triangle-to-square doodle featured in 2012.

Now at
PolyAI
Links
Google Scholar, X and LinkedIn

Timeline

  1. April 2012

    Triangle-to-square animation

    In April 2012, he creates a blog animation that encircles an equilateral triangle with squiggles until it turns into a square.

  2. 2013

    A Google PhD Fellowship

    After a BA in Mathematics at Cambridge and an MSc in Speech and Language Processing at the University of Edinburgh, he works toward a dialogue-systems PhD under Steve Young in Cambridge's Dialogue Systems Group. In July 2013, he is selected as a Global Google PhD Fellow, one of 39 students in the programme's fifth class.

  3. 2014

    Dialog state tracking papers

    In June 2014, he coauthors two SIGDIAL papers: The Second Dialog State Tracking Challenge with Blaise Thomson and Jason D. Williams, and Word-Based Dialog State Tracking with Recurrent Neural Networks with Thomson and Steve Young.

  4. 2017

    Smart Reply at Google

    At Google, he is first author of Efficient Natural Language Response Suggestion for Smart Reply in 2017, with Ray Kurzweil among the authors. The method uses feed-forward neural networks to encode messages and is tested in Inbox by Gmail, matching sequence-to-sequence quality at a small fraction of its computational requirements and latency.

  5. 2019

    Conversational datasets and ConveRT

    As VP Research at PolyAI, in April 2019, he is first author of A Repository of Conversational Datasets, which provides reproducible Reddit, OpenSubtitles, and Amazon QA datasets with a 1-of-100 response-selection accuracy test. In November 2019, he and coauthors including Nikola Mrkšić and Ivan Vulić introduce ConveRT, a lightweight dual-encoder pretraining framework for conversational tasks.

  6. April 2024

    Reka's multimodal models

    His career includes roles at Google, Apple, and Reka. In April 2024, he coauthors the Reka Team's report on Core, Flash, and Edge, multimodal language models trained from scratch to process text, images, video, and audio.

  7. August 2025

    Raven v2 at PolyAI

    In August 2025, as PolyAI's VP of Research, he introduces Raven v2, the company's in-house LLM for customer-service conversations. The model handles multi-turn flows and real-time function calls and serves multiple enterprise customers.

  8. As of October 2026

    VP of Research at PolyAI

    In November 2025, he discusses PolyAI's unified multilingual system for voice agents on its Deep Learning podcast. In July 2026, he introduces Dialog-RSN-1, an audio-native model that combines turn-taking, speech recognition, function calling, and response. As of October 2026, he is VP of Research at PolyAI and part of its founding research team with Nikola Mrkšić, Shawn Wen, and Eddy Su.

Career

  1. NowPolyAIVP of Research

Education

  1. University of CambridgeBA in Mathematics
  2. University of EdinburghMSc in Speech and Language Processing

From the biography.

  1. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. 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. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Google. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

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 Matthew Henderson yet. Send us a free-licensed photo.

  1. Squiggles, Spheres, and Taxes – Math Munch mathmunch.org. Checked 21 October 2026.
  2. Matthew Henderson receives 2013 Google PhD Fellowship | Cambridge Network cambridgenetwork.co.uk. Checked 21 October 2026.
  3. The Second Dialog State Tracking Challenge - ACL Anthology aclanthology.org. Checked 21 October 2026.
  4. Word-Based Dialog State Tracking with Recurrent Neural Networks - ACL Anthology aclanthology.org. Checked 21 October 2026.
  5. Matt Henderson matthen.com. Checked 21 October 2026.
  6. Efficient Natural Language Response Suggestion for Smart Reply research.google. Checked 21 October 2026.
  7. [1904.06472] A Repository of Conversational Datasets arxiv.org. Checked 21 October 2026.
  8. A Repository of Conversational Datasets | PolyAI poly.ai. Checked 21 October 2026.
  9. GitHub - PolyAI-LDN/conversational-datasets: Large datasets for conversational AI github.com. Checked 21 October 2026.
  10. [1911.03688] ConveRT: Efficient and Accurate Conversational Representations from Transformers arxiv.org. Checked 21 October 2026.
  11. [2404.12387] Reka Core, Flash, and Edge: A Series of Powerful Multimodal Language Models arxiv.org. Checked 21 October 2026.
  12. Raven v2: PolyAI's LLM Built for Customer Conversations poly.ai. Checked 21 October 2026.
  13. What does truly multilingual CX sound like? | PolyAI poly.ai. Checked 21 October 2026.
  14. Dialog-RSN-1: PolyAI's Audio-Native Dialog Model poly.ai. Checked 21 October 2026.
  15. About PolyAI: Leadership, Mission, and Company Story poly.ai. Checked 21 October 2026.