Applied Interpretability
Exploring how mechanistic interpretability techniques can reveal internal reasoning processes of security-focused language models.
Featured Projects
Selected projects and external publications.
LUCID (LLM-driven Understanding, Classification & Insight for Detections)
A framework for transforming traditional detection engineering logic into LLM-driven reasoning systems.
Defending at Machine Speed
Research exploring how LLMs can be guided with security context to accelerate threat detection and response workflows.
Toward Quantitative Modeling of Cybersecurity Risks due to AI Misuse
Proposing approaches to quantify emerging cybersecurity risks associated with AI misuse.