PeopleofSuperintelligence

Researcher

Wittawat Jitkrittum

Also known as วิทวัส จิตกฤตธรรม, Wit

Wittawat Jitkrittum is a machine learning researcher whose work on model routing and cascades makes large language models cheaper to run. After computer science studies at SIIT in Thailand and the Tokyo Institute of Technology, he earned a PhD in machine learning at the Gatsby Unit of University College London. His 2017 paper A Linear-Time Kernel Goodness-of-Fit Test won a NeurIPS Best Paper Award. He worked as a postdoctoral researcher at the Max Planck Institute for Intelligent Systems from 2018, then as a research scientist at Google Research from 2020, and his paper Faster Cascades via Speculative Decoding received an ICLR 2025 Outstanding Paper Honorable Mention. Since February 2026, he is an independent researcher working on model orchestration.

Timeline

  1. 2010–2012

    Computer science in Thailand and Tokyo

    He earns a BSc in computer science at SIIT in Thailand and an MEng in computer science at the Tokyo Institute of Technology. A full Okazaki Kaheita Scholarship supports his master's degree from 2010 to 2012.

  2. 2013–2017

    PhD at the Gatsby Unit

    A full scholarship from the Gatsby Unit funds his PhD in machine learning at University College London from 2013 to 2017. From 2015, his research develops sample-efficient, kernel-based statistical tests for comparing distributions and models.

  3. 2017

    A NeurIPS best paper award

    In 2017, A Linear-Time Kernel Goodness-of-Fit Test, written with Wenkai Xu, Zoltan Szabo, Kenji Fukumizu and Arthur Gretton, wins a NeurIPS Best Paper Award. The test learns the features that best indicate differences between samples and a reference model, at a computational cost linear in the number of samples.

  4. 2018–2020

    Postdoc at Max Planck

    From 2018 to 2020, he is a postdoctoral researcher at the Max Planck Institute for Intelligent Systems.

  5. 2020–2024

    Research scientist at Google Research

    He joins Google Research as a research scientist in 2020, and from 2022 his research covers model routing for efficient inference and teacher-based training of large language models. In 2023, he is first author of a NeurIPS paper with Neha Gupta, Aditya Krishna Menon, Harikrishna Narasimhan, Ankit Singh Rawat and Sanjiv Kumar that asks when confidence-based cascade deferral suffices.

  6. 2023–2024

    Two Google Tech Impact Awards

    He receives a Google Research Tech Impact Award in 2023 for his work on cascade models. In 2024, a Google Tech Impact Award for model routing names that project one of the ten most impactful of the year.

  7. 2025

    Speculative cascades and UniRoute

    Faster Cascades via Speculative Decoding, written with Harikrishna Narasimhan, Ankit Singh Rawat, Seungyeon Kim, Neha Gupta, Aditya Krishna Menon and Sanjiv Kumar, earns an Outstanding Paper Honorable Mention at ICLR 2025. Its speculative cascades implement a cascade's deferral rule through speculative execution and, on T5 and Gemma models, match the quality of a regular cascade at reduced inference cost. In February 2025, he also posts Universal Model Routing for Efficient LLM Inference, whose UniRoute method routes prompts among more than 30 unseen LLMs.

  8. Since February 2026

    Independent researcher on orchestration

    He serves as a senior research scientist at Google Research from 2024 and at Google DeepMind from 2025, then becomes an independent researcher in February 2026. His current research thread is model orchestration, which asks how many generalist and specialist models of various sizes can complete a task together.

Career

  1. 2025–2026Google DeepMindSenior Research Scientist
  2. 2020–2025Google ResearchResearch scientist, Senior Research Scientist
  3. 2018–2020Max Planck Institute for Intelligent SystemsPostdoctoral researcher

Education

  1. 2017University College LondonPhD in machine learning
  2. 2012Institute of Science TokyoMEng in computer science
  3. Sirindhorn International Institute of TechnologyBSc in computer science

From the biography.

Videos

YouTube videos in which Wittawat Jitkrittum speaks or appears, newest first. A model chose them from YouTube search results. List updated 21 October 2026.

  1. A Linear-Time Kernel Goodness-of-Fit Test - NIPS Best Paper '17The TWIML AI Podcast with Sam Charrington · 24 January 2018 · 24 min

Video data from YouTube API Services. By watching you agree to the YouTube Terms of Service. How we use YouTube data.

  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 Wittawat Jitkrittum yet. Send us a free-licensed photo.

  1. Wittawat Jitkrittum wittaw.at. Checked 20 October 2026.
  2. Awards (NeurIPS 2017) neurips.cc. Checked 20 October 2026.
  3. [1705.07673] A Linear-Time Kernel Goodness-of-Fit Test arxiv.org. Checked 20 October 2026.
  4. NeurIPS Best Paper: A Linear-Time Kernel Goodness-of-Fit Test | TWIML twimlai.com. Checked 20 October 2026.
  5. [2307.02764] When Does Confidence-Based Cascade Deferral Suffice? arxiv.org. Checked 20 October 2026.
  6. Announcing the Outstanding Paper Awards at ICLR 2025 – ICLR Blog blog.iclr.cc. Checked 20 October 2026.
  7. Faster Cascades via Speculative Decoding research.google. Checked 20 October 2026.
  8. [2502.08773] Universal Model Routing for Efficient LLM Inference arxiv.org. Checked 20 October 2026.