One of my recent projects, in collaboration with Prof. Peter A. Beerel at USC, explores secure and distributed inference for large language models. Our work proposes a lightweight framework that can detect and isolate faulty or malicious behavior during ML inference, a critical need for edge deployments. This project was accepted to GLSVLSI 2025, marking my first academic publication.
This is the link to our paper. Give it a read :)
Contributed to deep learning based tumor classification using MRI data, focusing on early-stage detection techniques. Gained exposure to clinical imaging workflows and interdisciplinary collaboration in a short term research capacity in Fan Magnetic Resonance (MR) Imaging Lab under Prof. Zhaoyang Fan.
Conducted a literature based research project on unified GPU architectures under Prof. Rahul Ratnakumar and presented applications and theoretical optimizations for parallel processing in healthcare and AI domains.