PeopleofSuperintelligence

Researcher

Boris Hanin

Boris Hanin is a mathematician who applies mathematical tools to questions in deep learning theory. He studied mathematics at Stanford and Northwestern, held an NSF postdoctoral fellowship in MIT's mathematics department, and moved from Texas A&M to Princeton. His research examines narrow ReLU networks, gradient stability in deep networks, compute-efficient transformers, and diffusion models. He contributed to The Principles of Deep Learning Theory and received a 2024 Sloan Research Fellowship in mathematics. As of October 2026, he is an associate professor at Princeton, where he studies deep learning, probability, and spectral asymptotics.

Timeline

  1. 2009–2014

    Mathematics at Stanford and Northwestern

    He earns a B.S. in mathematics at Stanford in 2009 and completes a Ph.D. in mathematics at Northwestern in 2014.

  2. 2014–2017

    MIT postdoc and Texas A&M

    After earning his mathematics Ph.D. at Northwestern in 2014, he is an NSF Postdoctoral Fellow in MIT's mathematics department. He joins the math faculty at Texas A&M as an assistant professor in 2017.

  3. 2017

    Narrow ReLU approximation

    In 2017, he proves that ReLU nets with width d+1 can approximate any continuous convex function of d variables and gives depth estimates for narrow-network approximation.

  4. 2018

    Random matrices and deep gradients

    In 2018, Hanin and Mihai Nica post Products of Many Large Random Matrices and Gradients in Deep Neural Networks, which quantifies exploding and vanishing gradients in randomly initialized deep ReLU networks.

  5. 2020

    Joining Princeton ORFE

    He joins Princeton's Department of Operations Research and Financial Engineering in 2020 as an assistant professor, following an MIT mathematics postdoc and an assistant professorship at Texas A&M.

  6. 2021

    The Principles of Deep Learning Theory

    He contributes research to The Principles of Deep Learning Theory with Daniel A. Roberts and Sho Yaida in 2021. The book develops an effective-theory approach to understanding practical deep neural networks.

  7. 2024

    A Sloan Research Fellowship

    The Alfred P. Sloan Foundation names him a 2024 Sloan Research Fellow in mathematics.

  8. 2024

    Applying physics to neural networks

    In 2024, he applies mathematical physics to identify principles that govern large neural networks and aims to help developers improve AI system designs before building them.

  9. 2025–2026

    Transformer training and diffusion models

    In 2025, he coauthors the NeurIPS paper Don't be lazy: CompleteP enables compute-efficient deep transformers, which studies how hyperparameter choices transfer as model size changes. In 2026, he and Peter Halmos connect score-based diffusion sampling to adiabatic transport of Schrödinger ground states.

  10. As of October 2026

    Associate Professor at Princeton

    In Spring 2025, Princeton Engineering commends him for teaching ORF 350. As of October 2026, he is an associate professor of operations research and financial engineering at Princeton, working on deep learning, probability, and spectral asymptotics.

Career

  1. NowPrinceton UniversityAssociate professor
  2. From 2017Texas A&M UniversityAssistant professor
  3. 2014–2017Massachusetts Institute of TechnologyNSF Postdoctoral Fellow

Education

  1. 2014Northwestern UniversityPh.D. in mathematics
  2. 2009Stanford UniversityB.S. in mathematics

From the biography.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Boris Hanin yet. Send us a free-licensed photo.

  1. Boris Hanin simons.berkeley.edu. Checked 26 October 2026.
  2. Three faculty members named Sloan Research Fellows engineering.princeton.edu. Checked 26 October 2026.
  3. [1708.02691] Universal Function Approximation by Deep Neural Nets with Bounded Width and ReLU Activations arxiv.org. Checked 26 October 2026.
  4. [1812.05994] Products of Many Large Random Matrices and Gradients in Deep Neural Networks arxiv.org. Checked 26 October 2026.
  5. Boris Hanin | boris-hanin boris-hanin.github.io. Checked 26 October 2026.
  6. [2106.10165] The Principles of Deep Learning Theory arxiv.org. Checked 26 October 2026.
  7. [2505.01618] Don't be lazy: CompleteP enables compute-efficient deep transformers arxiv.org. Checked 26 October 2026.
  8. [2606.05217] The Score Hamiltonian: Mapping Diffusion Models to Adiabatic Transport arxiv.org. Checked 26 October 2026.
  9. Faculty commended for outstanding teaching engineering.princeton.edu. Checked 26 October 2026.
  10. Boris Hanin - Princeton Engineering engineering.princeton.edu. Checked 26 October 2026.