Ananda Theertha Suresh Google Research, New York email: stheertha at gmail dot com [Google scholar]

I am a research scientist at Google Research, New York. I obtained my PhD in electrical and computer engineering from University of California, San Diego, where I was advised by Alon Orlitsky. Prior to joining UCSD, I obtained a Bachelor's degree in Engineering Physics from Indian Institute of Technology, Madras.

Research interests

My current research is primarily centered on developing algorithms and theoretical formulations for addressing resource and privacy considerations for machine learning systems in practice. Specific topics include federated learning, unlearning, privacy, and language models.

I am a theoretician by training and I have worked on basic statistical problems such as distribution estimation and testing during my PhD. I am always happy to discuss theoretical problems in the union of machine learning, information theory, and statistics.

Selected publications

Please see the Google scholar page for a full list of papers.

• SpecTr: Fast Speculative Decoding via Optimal Transport, NeurIPS 2023
  with Z. Sun, J. Ro, A. Beirami, H. Jain and F. Yu [pdf]

• Remember what you want to forget: Algorithms for machine unlearning, NeurIPS 2021
  with A. Sekhari, J. Acharya, and G. Kamath [pdf]

• On the Renyi Differential Privacy of the Shuffle Model, CCS 2021
  with A. Girgis, D. Data, S Diggavi, and P. Kairouz [pdf] (Best paper award)

• Optimal multiclass overfitting by sequence reconstruction from hamming queries, ALT 2020
  with J. Acharya [pdf] (Best paper award)

• Three approaches for personalization with applications to federated learning, Manuscript
  with Y. Mansour, M. Mohri, and J. Ro [pdf]

• Distributed mean estimation with limited communication, ICML 2017
  with F. Yu, H. B. McMahan, and S. Kumar [pdf]

• A unified maximum likelihood approach for optimal distribution property estimation, ICML 2017
  with J. Acharya, H. Das, and A. Orlitsky [pdf] (Best paper award honorable mention)

• Optimal prediction of the number of unseen species, PNAS 2016
  with A. Orlitsky and Y. Wu [pdf]

• Competitive distribution estimation: Why is Good-Turing good, NeurIPS 2015
  with A. Orlitsky [pdf] [talk] (Best paper award)