Shiva Kasiviswanathan

Shiva Kasiviswanathan

I am a Principal Applied Scientist at Amazon, working on the theory and practice of machine learning.

Shiva Kasiviswanathan speaking at the Agentic AI Summit
Principal Applied Scientist Amazon, Bay Area, California

Research interests

  • LLM post-training and alignment Reliable reasoning, evaluation, and adaptation for large language models.
  • Machine Learning Theory Foundations of learning, optimization, and reliable statistical methods.
  • Differential privacy Rigorous privacy guarantees for statistical and machine learning systems.

Selected publications

All research
  1. ICLR 2026

    Training Large Language Models to Reason in Parallel with Global Forking Tokens

    With S. Jia and X. Wang.

    Read the Amazon Science blog article about this work.

  2. ICML 2026

    Dependence-Aware Label Aggregation for LLM-as-a-Judge via Ising Models

    With K. Balasubramanian and A. Podkopaev.

    Read the Amazon Science blog article about this work.

  3. AISTATS 2019

    Subsampled Renyi Differential Privacy and Analytical Moments Accountant

    With Y. Wang and B. Balle.

    Notable Paper Award.

  4. COLT 2018

    Restricted Eigenvalue from Stable Rank with Applications to Sparse Linear Regression

    With M. Rudelson.

  5. SICOMP 2011

    What Can We Learn Privately?

    With H. K. Lee, K. Nissim, S. Raskhodnikova, and A. Smith.

  6. STOC 2010

    The Price of Privately Releasing Contingency Tables and the Spectra of Random Matrices with Correlated Rows

    With M. Rudelson, A. Smith, and J. Ullman.