Devvrit Khatri

devvrit [at] cs [dot] utexas [dot] edu

University of Texas, Austin

About Me

I am a founding member at Mirendil. I am also a fifth year PhD student at UT Austin, advised by Prof. Inderjit Dhillon. My research interests lie in Large Scale Machine Learning, with a focus on efficiency and scalability. I aim to build fundamental ML algorithms that exhibit strong empirical performance coupled with real-world deployability. In a quiet affirmation of their utility, many of these algorithms have been integrated into the workflows of Google, Meta, and Microsoft, among other organisations. Throughout my academic journey, I have worked at several industry labs. I worked at MSR in 2019 during my undergrad, have been a Student Researcher at Google from 2022-2025 (working closely with Prateek Jain), and was a Student Researcher at Meta (with Rishabh Agarwal).

In a fantastic previous life, I earned my Bachelor's from BITS Pilani, Goa, where I graduated with the Institute Bronze Medal. In my final year I worked at Microsoft Research India, where I had the good fortune of having Ravishankar Krishnaswamy as my undergraduate thesis advisor, and worked closely with Harsha Simhadri.

Latest Updates

April 2026	Joined Mirendil as a founding member 🚀
April 2026	Organizing SPOT Workshop@ICLR 26'!
Feb 2026	ScaleRL accepted as Oral at ICLR'26! 🎊
May-Dec 2025	Interned at Meta, Menlo Park with Rishabh Agarwal
September 2024	MatFormer accepted at NeurIPS'24! See you all in Vancouver 🎉
May 2024	Interning@GoogleDeepMind India with Prateek Jain.
January 2024	CASPR and Efficacy of Dual-Encoders accepted at ICLR'24. 💥
December 2023	MatFormer wins Best Paper Award at ENLSP@NeurIPS23 🏆
December 2023	Attending NeurIPS at NOLA. Come say hello 🤝
October 2023	MatFormer Preprint now available, check it out!
September 2023	SONew and Greedy Pruning accepted at NeurIPS'23. Wohoo!
May 2023	Interning@Google Research India with Prateek Jain. Say hello if you in Bangalore.

Selected Publications

arXiv

Learning, Fast and Slow: Towards LLMs That Adapt Continually

Rishabh Tiwari, Kusha Sareen, Lakshya A Agrawal, Joseph E. Gonzalez, Matei Zaharia, Kurt Keutzer, Inderjit S Dhillon, Rishabh Agarwal*, Devvrit Khatri*

Under Review

pdf BibTeX

@misc{fst,
      title={Learning, Fast and Slow: Towards LLMs That Adapt Continually}, 
      author={Rishabh Tiwari and Kusha Sareen and Lakshya A Agrawal and Joseph E. Gonzalez and Matei Zaharia and Kurt Keutzer and Inderjit S Dhillon and Rishabh Agarwal and Devvrit Khatri},
      year={2026},
      eprint={2605.12484},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2605.12484}, 
}

ICLR

The Art of Scaling Reinforcement Learning Compute for LLMs

Devvrit Khatri*, Lovish Madaan*, Rishabh Tiwari, Rachit Bansal, Sai Surya Duvvuri, Manzil Zaheer, Inderjit S. Dhillon, David Brandfonbrener, Rishabh Agarwal

ICLR, 2026 (Oral)

pdf BibTeX

@misc{khatri2025scalingrl,
      title={The Art of Scaling Reinforcement Learning Compute for LLMs}, 
      author={Devvrit Khatri and Lovish Madaan and Rishabh Tiwari and Rachit Bansal and Sai Surya Duvvuri and Manzil Zaheer and Inderjit S. Dhillon and David Brandfonbrener and Rishabh Agarwal},
      year={2025},
      eprint={2510.13786},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2510.13786}, 
}

NeurIPS

MatFormer: Nested Transformer for Elastic Inference

Devvrit*, Sneha Kudugunta*, Aditya Kusupati*, Tim Dettmers, Kaifeng Chen, Inderjit Dhillon, Yulia Tsvetkov, Hannaneh Hajishirzi, Sham Kakade, Ali Farhadi, Prateek Jain

NeurIPS, 2024; Also Best Paper@ENLSP, NeurIPS, 2023.

pdf BibTeX

@misc{devvrit2023matformer,
  title={MatFormer: Nested Transformer for Elastic Inference}, 
  author={Devvrit and Sneha Kudugunta and Aditya Kusupati and Tim Dettmers and Kaifeng Chen and Inderjit Dhillon and Yulia Tsvetkov and Hannaneh Hajishirzi and Sham Kakade and Ali Farhadi and Prateek Jain},
  year={2023},
  eprint={2310.07707},
  archivePrefix={arXiv},
  primaryClass={cs.LG}
}

ICLR

Combining Axes Preconditioners through Kronecker Approximation for Deep Learning

Sai Surya Duvvuri, Devvrit, Rohan Anil, Cho-Jui Hsieh, Inderjit S Dhillon

ICLR, 2024

pdf BibTeX

@inproceedings{
  anonymous2024combining,
  title={Combining Axes Preconditioners through Kronecker Approximation for Deep Learning},
  author={Anonymous},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024},
  url={https://openreview.net/forum?id=8j9hz8DVi8}
  }

NeurIPS

S3GC: Scalable Self-Supervised Graph Clustering

Devvrit*, Aditya Sinha, Inderjit Dhillon, Prateek Jain

NeurIPS, 2022

pdf BibTeX

@inproceedings{
  devvrit2022sgc,
  title={S3{GC}: Scalable Self-Supervised Graph Clustering},
  author={Fnu Devvrit and Aditya Sinha and Inderjit S Dhillon and Prateek Jain},
  booktitle={Advances in Neural Information Processing Systems},
  editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
  year={2022},
  url={https://openreview.net/forum?id=ldl2V3vLZ5}
  }

NeurIPS

DiskANN: Fast Accurate Billion-point Nearest Neighbor Search on a Single Node

Suhas Jayaram Subramanya, Devvrit, Rohan Kadekodi, Ravishankar Krishaswamy, Harsha Vardhan Simhadri

NeurIPS 2019

pdf BibTeX

@inproceedings{NEURIPS2019_09853c7f,
  author = {Jayaram Subramanya, Suhas and Devvrit, Fnu and Simhadri, Harsha Vardhan and Krishnawamy, Ravishankar and Kadekodi, Rohan},
  booktitle = {Advances in Neural Information Processing Systems},
  editor = {H. Wallach and H. Larochelle and A. Beygelzimer and F. d\textquotesingle Alch\'{e}-Buc and E. Fox and R. Garnett},
  pages = {},
  publisher = {Curran Associates, Inc.},
  title = {DiskANN: Fast Accurate Billion-point Nearest Neighbor Search on a Single Node},
  url = {https://proceedings.neurips.cc/paper_files/paper/2019/file/09853c7fb1d3f8ee67a61b6bf4a7f8e6-Paper.pdf},
  volume = {32},
  year = {2019}
  }