HSF-India Trainee Fellow: Krishna Bulchandani
Fellowship dates: Aug – Oct, 2025
Home Institution:
Project: Prototyping Large Language Models for CMS Event Analysis in the HL-LHC Era
My project investigates transformer-based approaches for track reconstruction in CMS Phase-2 at the High-Luminosity LHC, where data volumes and pile-up conditions are unprecedented. By representing detector hits as doublets—pairs of geometrically consistent hits—I aim to reduce input complexity and apply efficient transformer architectures with scalable attention mechanisms such as locality-sensitive hashing (LSH). The model will be pretrained with masked token objectives and fine-tuned for full track reconstruction in high pile-up environments. The goal is to deliver a compact, CMS Phase-2–ready transformer pipeline that matches or surpasses current graph neural network trackers in efficiency, latency, and scalability, paving the way for modern large-model methods in future particle physics reconstruction tasks.More information: My project proposal
Mentors:
-
David Lange (Princeton University)
-
Peter Elmer (Princeton University)
-
Liv Våge (Princeton University)
Current Status
Contact me: