Portrait of Siddhant Chaudhary

Siddhant Chaudhary | सिद्धांत चौधरी

CS Grad @ Kahlert School of Computing, University of Utah

About

Hello! I'm a second-year CS PhD student at the Kahlert School of Computing, University of Utah. I'm fortunate to be working with Prof. Aditya Bhaskara. I'm also a part of Utah's Theory Group. Last summer, I spent my time at the Google Bengaluru office, where I worked on building an end-to-end auto-configurator for vLLM inference on TPUs. Before that, I spent a year as a research assistant at the Laboratory for Computational Social Systems (LCS2) @IIT Delhi (led by Dr. Tanmoy Chakraborty) wherein I worked on LLM and KV cache compression algorithms, and also as a research assistant at the Networks and Learning Group at TIFR Mumbai (led by Dr. Abhishek Sinha) wherein I worked on problems in online learning and fairness in contextual bandits. During my masters at CMI, I was also fortunate to be advised by Prof. K. V. Subrahmanyam, wherein we explored the complexity of various RL-based policy iteration algorithms on the hypercube.

I'm broadly interested in fundamental problems in artificial intelligence with a particular focus on robustness and efficiency, which span both theory and systems. On the theory side, I'm currently working on designing efficient algorithms for fundamental estimation problems where the source of data is non-IID (particularly, fundamental estimation problems in the heteroskedastic setting). On the systems side, I'm exploring algorithmic and systems-level optimization problems in model inference. This includes things like KV cache and model compression. I also love keeping in touch with the open source inference ecosystem, including vLLM, TensorRT and llama.cpp.

The best way to reach me is via my email (urssidd@gmail.com), and you can find my other links at the bottom of this page. I'm always keen on collaborating with people on problems of interest!

Theory interests: High-dimensional statistics, Optimization, Applied Probability.

ML Systems interests: Efficient Inference, Model Compression, XPUs

Other interests: Sleeping, Eating, Progressive Metal, Lifting, Speed Typing, Gaming, Investing

Publications


Note: * represents equal contribution

Industry/OSS Experience

  • PhD SWE Intern, Google
    May 2026 - August 2026
    Efficient inference in LLMs.
  • Open Source Developer, Google Summer of Code
    May 2024 – October 2024
    Ported ColBERT to Julia; developed and maintained ColBERT.jl.
  • Open Source Developer, Google Summer of Code
    May 2022 - October 2022
    Contributed to and wrote various packages in the JuliaStats ecosystem.

News

  • Sep 2026 Our second work on efficient algorithms for linear regression with hetereskedastic errors accepted at NeurIPS 2026!
  • June 2026 Our work on designing new algorithms for linear regression with hetereskedastic variances accepted at UAI 2026!
  • May 2026 Started as an intern at Google! Working on efficient inference in vLLM on TPUs.
  • Aug 2025 Our paper on a new KV cache compression technique accepted at NeurIPS 2025!
  • Aug 2025 Started my PhD in computer science at KSoC, U of U!
  • Apr 2025 My work on ColBERT.jl accepted as a main talk to JuliaCon 2025!
  • Jan 2025 Our paper on PruneNet, a novel structured model compression technique, accepted to ICLR 2025!
  • Jun 2024 Our paper on fair contextual bandits accepted at FoRLaC@ICML2024!
  • May 2024 Joined LCS2@IITD as an RA!
  • May 2024 Graduated from CMI with a BS + MS in Computer Science!