PeopleofSuperintelligence

Researcher

Abhishek Kumar

Abhishek Kumar is a machine learning researcher whose work spans multi-view clustering, generative models and language models. He earned a computer science PhD at Maryland under Hal Daumé III and worked at IBM before Google. His research includes BlockDrop, score-based diffusion, language-model self-training, training-free diffusion personalization and causal representation learning. By April 2026, he is with Amazon after Google DeepMind.

Links
Website and Google Scholar

Timeline

  1. 2011

    Co-regularized spectral clustering

    In 2011, he coauthors Co-regularized Multi-view Spectral Clustering with Piyush Rai and Hal Daumé III for NIPS 2011. The method co-regularizes clustering hypotheses across multiple views of the data.

  2. 2012

    Maryland PhD and IBM research

    He earns a computer science PhD at the University of Maryland in 2012 under Hal Daumé III, with a dissertation titled Learning with Multiple Similarities. During his master's at CRSS, he studies robust automatic speech recognition under John H. L. Hansen, then works as a researcher at IBM's TJ Watson Research Center in Yorktown Heights before Google.

  3. 2017–2018

    BlockDrop and disentangled latents

    In 2017, he coauthors BlockDrop with Zuxuan Wu, Tushar Nagarajan, Steven Rennie, Larry S. Davis, Kristen Grauman and Rogerio Feris. A policy network selects which residual blocks a ResNet runs for each image. He also coauthors Variational Inference of Disentangled Latent Concepts from Unlabeled Observations with Prasanna Sattigeri and Avinash Balakrishnan; its ICLR 2018 version uses variational inference to learn disentangled latent factors from unlabeled observations.

  4. 2020–2021

    Score-based modeling wins at ICLR

    In November 2020, he coauthors Score-Based Generative Modeling through Stochastic Differential Equations, which gradually turns data into noise and uses a reverse-time SDE to generate data. The paper wins one of ICLR's eight 2021 Outstanding Paper Awards.

  5. 2024

    Self-training for language models

    In 2024, he coauthors Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models. The team tests an expectation-maximization self-training method with binary feedback on mathematical reasoning and code generation.

  6. 2024

    RB-Modulation and language models

    In May 2024, he coauthors RB-Modulation, a training-free method for personalizing diffusion models. He works as a research scientist at Google Brain, where his work spans transformer attention biases, language-model efficiency through distillation and activation sparsity, and alternative architectures for sequence modeling.

  7. 2025

    Causal representation learning

    In 2025, he coauthors Score-based Causal Representation Learning: Linear and General Transformations, a JMLR paper that studies recovery of latent causes and graphs from interventional data.

  8. April 2026

    Amazon affiliation

    By April 2026, he is with Amazon after Google DeepMind.

Career

  1. Google DeepMindResearch Scientist
  2. IBM ResearchResearcher

Education

  1. University of MarylandPhD
  2. CRSSMaster's

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Abhishek Kumar. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  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 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 Abhishek Kumar yet. Send us a free-licensed photo.

  1. Co-regularized Multi-view Spectral Clustering papers.nips.cc. Checked 22 October 2026.
  2. Hal Daumé III - advisees users.umiacs.umd.edu. Checked 22 October 2026.
  3. Abhishek Kumar abhishek.ml. Checked 22 October 2026.
  4. BlockDrop: Dynamic Inference Paths in Residual Networks arxiv.org. Checked 22 October 2026.
  5. Variational Inference of Disentangled Latent Concepts from Unlabeled Observations arxiv.org. Checked 22 October 2026.
  6. Score-Based Generative Modeling through Stochastic Differential Equations arxiv.org. Checked 22 October 2026.
  7. 8 Outstanding Papers At ICLR 2021 analyticsindiamag.com. Checked 22 October 2026.
  8. Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models arxiv.org. Checked 22 October 2026.
  9. RB-Modulation: Training-Free Personalization of Diffusion Models using Stochastic Optimal Control arxiv.org. Checked 22 October 2026.
  10. Score-based Causal Representation Learning: Linear and General Transformations jmlr.org. Checked 22 October 2026.
  11. Learning world models from doing and seeing: Causal representation learning and applications to robot pose estimation web.tcs.tifr.res.in. Checked 22 October 2026.