Presented our CCTA-RADS poster at ACL 2026 in San Diego
Sudharshan Balaji சுதர்ஷன் பாலாஜி
pronounced SOO-dhahr-shun buh-LAH-jee
@bsudharshan2001
Ph.D. researcher at University of South Florida, working in the SPRAI Lab with Dr. Ning Wang. My current work focuses on AI safety and alignment.
I work on large language models, unlearning, and agentic systems: how they retain information, follow intent, fail under pressure, and can be made more reliable.
Before the Ph.D., I finished my Master's at USF and worked across security, federated learning, and applied NLP. Outside of research, I watch football (the European kind), catch F1 races, and keep a few side projects alive.
Recent
Our new preprint, "One Modality to Forget Them All", is now on arXiv
Our paper, Snail: Scheduler-Driven Backdoor Injection in Split Federated Learning via Gradual Perturbation, was accepted to IEEE CNS 2026; I am the second author
Paper "Learning from Textual Radiology Reports: A Benchmark Dataset for Coronary CT Angiography" accepted to the ACL 2026 Industry Track
Started PhD continuing under Dr. Ning Wang
Graduated with Master's in CS from USF
Thesis on LLMs for Network Intrusion Detection published to ProQuest
Defended thesis successfully
Started TA'ing "Security & Privacy in ML"
Joined SPRAI Lab with Dr. Ning Wang