Research Assistant
University of Maryland, Baltimore County
Aug 2026 - Present
Researching how large language models respond to certain classes of security vulnerabilities.
Testing different prompting styles across multiple LLMs under controlled conditions (temperature settings, model type, and context-window length) to compare how each model reacts.
Comparing different data-flow techniques, i.e. how input is structured and passed to the model, to see how they change model behaviour.
Large Language Models · AI Security · Prompt Engineering · LLM Evaluation






