CV

Sanjit Dandapanthula

ML researcher @ a new AI lab

About

I'm currently a member of technical staff at a new AI lab.

Previously, I was a PhD student at Carnegie Mellon University studying statistics and machine learning. During my time there, I also worked as a research intern at Abridge AI in New York City and Amazon Web Services in Santa Clara.

I studied mathematics and computer science at UCLA.

If you would like to contact me, please use one of the links on this page. Alternatively, you can try shouting my name in a public place.

Research

I was extremely lucky to be advised by Aaditya Ramdas (now at Stanford) and Nicholas Boffi at CMU. My research interests are in generative modeling, deep learning theory, and optimal transport.

I am most fascinated by theoretical questions with immediate practical implications, such as:

Why do modern machine learning methods work so well?
Where do they fall short, and how can we design better ones?

Beyond research

In my spare time, I've been known to lift heavy things, watch TV shows, read books, and play board games with friends. I've recently enjoyed playing Gloomhaven: Jaws of the Lion and highly recommend the Mistborn and The Stormlight Archive series by Brandon Sanderson. I also play Indian classical music on the Bansuri flute.

In the past, I ran marathons, biked long distance, and hiked glacier mountains supporting Asha for Education (please donate here!), and volunteered for the Crisis Text Line as a Crisis Counselor (donate here!).

Papers

Current as of November 2, 2026 — also see Google Scholar.

1. 01

   How to post-train on a surrogate: Envelope sampling mitigates reward hacking

   Sanjit Dandapanthula, Shuvom Sadhuka, Samir Khan, Michael Oberst, Aaditya Ramdas, and Alexandra Chouldechova.
   arXiv preprint, 2026.

   arXiv

2. 02

   Are we really tilting? The mechanics of reward guidance in flow and diffusion models

   Sanjit Dandapanthula and Nicholas M. Boffi.
   NeurIPS, 2026.

   arXiv code website proceedings

3. 03

   Downscaling land surface temperature data using edge detection and block-diagonal Gaussian process regression

   Sanjit Dandapanthula, Margaret Johnson, Madeleine Pascolini-Campbell, Glynn Hulley, and Mikael Kuusela.
   Environmental Data Science, 2026.

   arXiv code proceedings

4. 04

   Gradient descent dynamics for deep equilibrium models

   Sanjit Dandapanthula and Aaditya Ramdas.
   arXiv preprint, 2025.

   arXiv code

5. 05

   Optimal transportation and alignment between Gaussian measures

   Sanjit Dandapanthula, Aleksandr Podkopaev, Shiva Kasiviswanathan, Aaditya Ramdas, and Ziv Goldfeld.
   arXiv preprint, 2025.

   arXiv code

6. 06

   Offline changepoint localization using a matrix of conformal p-values

   Sanjit Dandapanthula and Aaditya Ramdas.
   Transactions on Machine Learning Research, 2026.

   arXiv code proceedings

7. 07

   Anytime-valid FDR control with the stopped e-BH procedure

   Hongjian Wang, Sanjit Dandapanthula, and Aaditya Ramdas.
   Statistics and Probability Letters, 2025.

   arXiv proceedings

8. 08

   Multiple testing in multi-stream sequential change detection

   Sanjit Dandapanthula and Aaditya Ramdas.
   arXiv preprint, 2025.

   arXiv code