PeopleofSuperintelligence

Researcher

Subhabrata Sen

Subhabrata Sen is a statistician whose research explains how algorithms recover structure from sparse graphs and high-dimensional data. He earned a statistics Ph.D. at Stanford in 2017 and studied community detection, contextual network models, and Bayesian regression. His work extends to neural-network pretraining and transfer learning, with a 2025 analysis of their effect on sample complexity. Harvard names him a Roslyn Abramson Award recipient in 2024 and lists him among its faculty engaged in AI as of October 2026.

Timeline

  1. 2015

    Sparse graph community detection

    With Andrea Montanari, he studies community detection in sparse random graphs through a semidefinite program. The 2015 paper shows that this program reaches the information-theoretic detection threshold at large bounded degree.

  2. 2016

    Optimization on sparse hypergraphs

    In 2016, he studies optimization on sparse random hypergraphs by comparing these problems with Gaussian optimization. The paper derives values for Max q-cut, Max XORSAT, and minimum bisection in the stochastic block model.

  3. 2017

    A Stanford statistics doctorate

    He earns a Ph.D. in statistics at Stanford in 2017, with Amir Dembo and Andrea Montanari as joint advisers.

  4. 2018

    Contextual Stochastic Block Models

    With Yash Deshpande, Andrea Montanari, and Elchanan Mossel, he studies inference of community structure from a sparse graph with high-dimensional node covariates. Their contextual stochastic-block-model analysis appears at NeurIPS in 2018.

  5. 2022

    Variational inference for regression

    With Sumit Mukherjee, he publishes “Variational Inference in high-dimensional linear regression” in the Journal of Machine Learning Research in 2022. The paper gives sufficient conditions for a naive mean-field approximation to the posterior log-normalizing constant and a variational formula for that constant.

  6. 2024

    The Roslyn Abramson Award

    He receives Harvard’s Roslyn Abramson Award in 2024 for excellence and sensitivity in teaching undergraduates.

  7. July 2025

    Provable benefits of pretraining

    In July 2025, he and Taj Jones-McCormick and Aukosh Jagannath publish an ICML analysis of unsupervised pretraining and transfer learning for a single-layer neural network. They show these approaches can reduce the sample complexity of high-dimensional supervised learning.

  8. As of October 2026

    Harvard faculty in AI

    Harvard lists him among its Faculty of Arts and Sciences members engaged in artificial intelligence. In 2026, he coauthors a survey of progress in high-dimensional statistics, connecting it with optimization, random matrix theory, information theory, and theoretical computer science.

Career

  1. NowHarvard UniversityFaculty member

Education

  1. 2017Stanford UniversityPhD in Statistics

From the biography.

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 Subhabrata Sen yet. Send us a free-licensed photo.

  1. Semidefinite Programs on Sparse Random Graphs and their Application to Community Detection arxiv.org. Checked 27 October 2026.
  2. Optimization on Sparse Random Hypergraphs and Spin Glasses arxiv.org. Checked 27 October 2026.
  3. Subhabrata Sen: Modeling & Statistical Analysis datascience.ucsd.edu. Checked 27 October 2026.
  4. Contextual Stochastic Block Models papers.nips.cc. Checked 27 October 2026.
  5. Variational Inference in high-dimensional linear regression jmlr.org. Checked 27 October 2026.
  6. Professorships and Awards oue.fas.harvard.edu. Checked 27 October 2026.
  7. Provable Benefits of Unsupervised Pre-training and Transfer Learning via Single-Index Models proceedings.mlr.press. Checked 27 October 2026.
  8. Faculty - Generative AI @ Harvard harvard.edu. Checked 27 October 2026.
  9. High-Dimensional Statistics: Reflections on Progress and Open Problems arxiv.org. Checked 27 October 2026.