Sudhanshu Agrawal
Hi everyone! My name is Sudhanshu, and I'm a Computer Science Master's student at Stanford. My current research interests are in methods to study the internal representations of large generative models, and in developing foundation models across modalities, in particular for the physical sciences and medicine.
Previously, I was an ML Research Engineer at Qualcomm AI, working on LLM efficiency research supervised by Mingu Lee. At Qualcomm, we worked on inventing new speculative decoding algorithms and architectures for edge applications — making LLMs fast enough to run locally on your phone or laptop!
I graduated from UCLA in 2023 with a double major in Computer Science and Mathematics. While I was at UCLA, I was fortunate to conduct research with Professor Aditya Grover on generative modeling and with Professor Levon Nurbekyan and Professor Samy Wu Fung on mean-field games.
In my free time, I like to surf, read, play the guitar, and sing. Feel free to reach out if you'd like to chat!
Experience
ML Research Engineer
ML Engineering Intern
ML Engineering Intern
ML and Data Science Intern
ML Intern
Education
Master of Science, Computer Science
Bachelor of Science, Computer Science
Bachelor of Science, Mathematics
ISC 12th Grade
Surfing the Activation Wave: Analyzing the Information and Geometry of LLM Hidden States for Speculative Decoding
Structuring The Future: Diffusion LLM Speculative Decoding via Calibrated Draft Graphs
A Comparative analysis of Layer-wise Representational Capacity in AR and Diffusion LLMs
VOCABTRIM: Vocabulary Pruning for Efficient Speculative Decoding in LLMs.
AdaEDL: Early Draft Stopping for Speculative Decoding of Large Language Models via an Entropy-based Lower Bound on Token Acceptance Probability
ExPT: Synthetic Pretraining for Few-Shot Experimental Design
Random Features for High-Dimensional Nonlocal Mean-Field Games
Patents
The following patent applications relate to 12 distinct inventions. Each invention has multiple US and global pending patents.
- US 19/782,156(lead inventor)
- TW 115121889(lead inventor)
- WO PCT/US2026/030830(lead inventor)
- WO PCT/US2026/030887(lead inventor)
- US 64/074,098(lead inventor)
- US 64/072,715(lead inventor)
- US 19/662,922
- US 19/634,806
- US 19/571,826(lead inventor)
- US 19/554,427(lead inventor)
- TW 115104375
- WO PCT/US2026/013469
- US 63/964,912
- US 63/944,624 (lead inventor)
- US 63/938,500 (lead inventor)
- WO PCT/CN2025/134909
- WO PCT/CN2025/124672
- US 63/872,751
- US 63/849,613 (lead inventor)
- US 19/273,664 (lead inventor)
- WO PCT/US2025/037170 (lead inventor)
- US 19/086,578 (lead inventor)
- US 18/983,103 (lead inventor)
- US 63/688,654 (lead inventor)
Blog
Generative AI for Experimental Design
Simulating The Motion of Charged Bodies
Invited Talks | Judgeships | Reviewing
- Reviewer: 2026 NeurIPS AXIOM Workshop on Foundations of Efficient Deep Learning
- Reviewer: 2026 ICML Workshop on the Foundations of Generative Models
- Reviewer: 2026 Transactions on Pattern Analysis and Machine Intelligence
- Reviewer: 2026 AAAI Conference
- Judge: 2025 UCSD Graduate Student Research Exposition
- Judge: 2025 San Diego State University Student Research Symposium
- Reviewer: 2025 ICML Efficient Systems for Foundation Models Workshop
- Reviewer: 2025 NeurIPS Efficient Natural Language and Speech Processing Workshop
- Speaker: 2024 UCSD and Qualcomm Graduate Students Tech Talk and Recruitment Event
- Speaker: 2024 UCSD, IEEE, Qualcomm Careers Panel
- Speaker: 2024 UCLA Mathematics Department Alumni Panel