Sriram Sai Ganesh

I am a Master’s student in Computer Science (HLT) at Johns Hopkins University, advised by Dr. Jason Eisner at the Center for Language and Speech Processing.

Broadly, I am interested in generalizable methods for compute-optimal machine learning. Attention as a primitive has been a powerful tool for sequence modeling in language and vision tasks; can we utilize this expressive power without paying a cost that is quadratic in sequence length?

Currently, I am working on a family of transformer decoder architectures that will allow for much faster inference. Improved throughput enables faster test-time reasoning, speculative decoding, and lower-latency agentic workflows, among other applications.

Before Johns Hopkins, I earned my B.S. at The Ohio State University, where I was fortunate to be advised by Dr. Srinivasan Parthasarathy. I studied efficient language and vision models while working on my thesis, including our work optimizing the Segment Anything Model for edge inference for disaster-response information retrieval.

news

Sep 16, 2025	Crisis Observatory has been accepted at ICDM ‘25.
Aug 26, 2025	Starting my MSE in Human Language Technologies at Johns Hopkins 🐦
May 05, 2025	Graduated with a B.S. in C.S.E. from The Ohio State University. Go Bucks!
Apr 09, 2025	Defended my undergrad thesis.
Jun 04, 2024	Awarded the URS scholarship for my honors thesis.
May 23, 2024	CPC@OSU is headed to the ICPC North American Championship. See y’all in Orlando!

publications

ICDM 2025

Crisis Observatory: Extracting Credible Signals During a Crisis in the Age of LLMs

K. Lo, P. Maneriker, S. Sai Ganesh, and 8 more authors

IEEE International Conference on Data Mining, Washington, D.C., Nov 2025

Abs PDF Poster

Systems for crisis response have required several different models for the analysis of unstructured text, such as identifying needs, locations, topics, routing, and matching of needs with available responders. Large Language Models (LLMs) have replaced task-specific models across various language processing tasks. However, LLMs are known to be limited by their training data, collected before the crisis. In this demo, we explore the use of LLMs for crisis response scenarios with rapidly evolving information environments. We show how augmentation of these models with external reliable sources of crisis-specific information can help build adaptive systems for response.