PeopleofSuperintelligence

Researcher

Kulin Shah

Kulin Shah is a machine learning researcher whose work studies generative models and language-model reasoning. He studied computer science at IIIT-Hyderabad, earned a BTech (Hons), and worked as a research fellow at Microsoft Research India before doctoral study at UT Austin. His earlier papers examined fair classification, normalizing flows, and Gaussian mixtures; at Google, he studied language-model reasoning and coauthored a Sudoku-solving study. In 2025, he coauthored an ICML Outstanding Paper on masked diffusion and a separate study of masked diffusion for generative recommendations, and served as a Snap Research intern. In 2026, he coauthored a paper on diffusion language models; as of October 2026, he is an ML Researcher at Apple Machine Learning Research.

Now at
Apple
Badges

Timeline

  1. 2017–2019

    Computer science at IIIT Hyderabad

    At IIIT-Hyderabad, he studies computer science, completes a B.Tech. (Hons.), and works with Naresh Manwani. His honours study there spans 2017 to 2019.

  2. 2021

    Fair adaptation and a fellowship

    In 2021, he coauthors Rawlsian Fair Adaptation of Deep Learning Classifiers, which proposes a threshold rule that minimizes error for the worst-off sensitive subpopulation. Before joining the Computer Science Ph.D. program at UT Austin, he works as a research fellow at Microsoft Research India with Amit Deshpande and Navin Goyal; Adam Klivans advises his doctoral study.

  3. 2022–2023

    Normalizing flows and Gaussian mixtures

    In 2022, he coauthors Learning and Generalization in Overparameterized Normalizing Flows, a theoretical and empirical study of flow training and generalization. It finds that overparameterization can hurt training in constrained flows and proves that unconstrained flows can learn distributions under minimal assumptions. In 2023, he coauthors Learning Mixtures of Gaussians Using the DDPM Objective, proving that gradient descent can efficiently recover mixture parameters under separation conditions.

  4. June 2023–March 2024

    Google Research and Sudoku reasoning

    At Google, he works with Rina Panigrahy on understanding and improving language-model reasoning. In 2024, he coauthors a study showing that Transformers trained on logical solver steps solve 94.21% of its Sudoku puzzles correctly.

  5. 2025

    An ICML Outstanding Paper

    In 2025, he coauthors Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions with Jaeyeon Kim, Vasilis Kontonis, Sham Kakade and Sitan Chen. It wins an ICML Outstanding Paper Award; adaptive token ordering raises Sudoku accuracy in pretrained masked diffusion models from under 7% to about 90%.

  6. 2025

    Masked diffusion for recommendations

    In 2025, he interns at Snap Research. In November 2025, he coauthors Masked Diffusion for Generative Recommendation, which uses masked diffusion to learn sequences of semantic IDs and predict several IDs in parallel.

  7. 2026

    Diffusion language model policy gradients

    In 2026, he coauthors Mask-Aware Policy Gradients for Diffusion Language Models.

  8. As of October 2026

    ML researcher at Apple

    As of October 2026, he is an ML Researcher at Apple Machine Learning Research.

Career

  1. NowAppleML Researcher
  2. 2025SnapIntern
  3. 2023–2024Google Research
  4. Microsoft ResearchResearch fellow

Education

  1. 2019International Institute of Information Technology, HyderabadB.Tech. (Hons.)
  2. University of Texas at AustinPh.D.

From the biography.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Kulin Shah. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  2. 16% of AI career moves in the 2020s went from a university to a companyIncludes Kulin Shah. 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.
  3. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Apple. 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.
  4. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Apple. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.

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 Kulin Shah yet. Send us a free-licensed photo.

  1. Kulin Shah kulinshah98.github.io. Checked 21 October 2026.
  2. Naresh Manwani - Students sites.google.com. Checked 21 October 2026.
  3. Rawlsian Fair Adaptation of Deep Learning Classifiers arxiv.org. Checked 21 October 2026.
  4. Learning and Generalization in Overparameterized Normalizing Flows proceedings.mlr.press. Checked 21 October 2026.
  5. Learning Mixtures of Gaussians Using the DDPM Objective arxiv.org. Checked 21 October 2026.
  6. Causal Language Modeling Can Elicit Search and Reasoning Capabilities on Logic Puzzles arxiv.org. Checked 21 October 2026.
  7. ICML 2025 Awards icml.cc. Checked 21 October 2026.
  8. Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions arxiv.org. Checked 21 October 2026.
  9. cv | Neil Shah nshah.net. Checked 21 October 2026.
  10. Masked Diffusion for Generative Recommendation arxiv.org. Checked 21 October 2026.