PeopleofSuperintelligence

Researcher

Ankit Singh Rawat

Ankit Singh Rawat studies methods that make language-model inference more efficient. He earned degrees at IIT Kanpur and UT Austin and held postdoctoral appointments at Carnegie Mellon, MIT and UMass Amherst. His papers explore confidence-based cascades, draft-model distillation and token-level uncertainty for language models. In 2025, his speculative-cascade paper received an ICLR Outstanding Paper Honorable Mention. As of October 2026, he is a Research Scientist at Google Research in New York.

Badges

Timeline

  1. 2010–2015

    Degrees in Kanpur and Austin

    He earns a B.Tech. in electrical engineering at IIT Kanpur in 2010 and a Ph.D. in electrical and computer engineering at UT Austin in 2015.

  2. 2015–2017

    Postdocs at CMU, MIT and UMass

    He is a postdoctoral fellow in Carnegie Mellon's Computer Science Department in 2015–2016 and a joint postdoc at MIT and UMass Amherst by October 2017.

  3. 2023

    Confidence-based cascade deferral

    In 2023, he coauthors When Does Confidence-Based Cascade Deferral Suffice?, which analyzes when confidence-based deferral fails and studies post-hoc rules that can improve accuracy-cost trade-offs.

  4. 2023

    DistillSpec for faster decoding

    Through DistillSpec in 2023, he uses knowledge distillation to align a small draft model with a larger target model; the method reports 10–45% speedups over standard speculative decoding on benchmarks.

  5. 2024

    Token-level uncertainty for LMs

    In 2024, he coauthors Language Model Cascades: Token-level uncertainty and beyond, a study showing that learned deferral rules using token-level uncertainty can improve cost-quality trade-offs for language-model tasks.

  6. 2024–2025

    Faster Cascades at ICLR

    In 2024, he coauthors Faster Cascades via Speculative Decoding, which applies a cascade's deferral rule through speculative execution. In 2025, the paper receives an ICLR Outstanding Paper Honorable Mention.

  7. 2025

    Gating is Weighting at COLM

    In 2025, he coauthors Gating is Weighting: Understanding Gated Linear Attention through In-context Learning, a COLM study of gating in linear-attention models. At NeurIPS 2025, he serves as a Senior Area Chair.

  8. As of October 2026

    Google Research in New York

    As of October 2026, he is a Research Scientist at Google Research in New York, with research interests in large-scale machine learning, coding theory and information theory.

Career

  1. NowGoogle ResearchResearch Scientist
  2. Until 2017Massachusetts Institute of TechnologyPostdoc
  3. Until 2017University of Massachusetts AmherstPostdoc
  4. 2015–2016Carnegie Mellon UniversityPostdoctoral fellow

Education

  1. 2015University of Texas at AustinPh.D.
  2. 2010Indian Institute of Technology KanpurB.Tech.

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Ankit Singh Rawat. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. 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.
  3. 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.
  4. 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.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Ankit Singh Rawat yet. Send us a free-licensed photo.

  1. Fall 2017 SIP Seminars – INSPIRE Lab | Rutgers University inspirelab.us. Checked 20 October 2026.
  2. When Does Confidence-Based Cascade Deferral Suffice? arxiv.org. Checked 20 October 2026.
  3. DistillSpec: Improving Speculative Decoding via Knowledge Distillation arxiv.org. Checked 20 October 2026.
  4. Language Model Cascades: Token-level uncertainty and beyond arxiv.org. Checked 20 October 2026.
  5. Faster Cascades via Speculative Decoding arxiv.org. Checked 20 October 2026.
  6. Announcing the Outstanding Paper Awards at ICLR 2025 blog.iclr.cc. Checked 20 October 2026.
  7. Publications – Wittawat Jitkrittum wittaw.at. Checked 20 October 2026.
  8. Ankit Singh Rawat research.google. Checked 20 October 2026.
  9. Google at NeurIPS 2025 research.google. Checked 20 October 2026.
  10. Ankit Singh Rawat ankitsrawat.github.io. Checked 20 October 2026.