Software security is the core reason I became a researcher. To me, a vulnerability is rarely just a technical failure — it is a symptom of how people and technology actually interact. Over more than 10 years I have worked that problem from many angles, from architecting financial backends serving millions of users to building threat detection for Google Threat Intelligence.

My PhD looked at why developers write vulnerable code; at FireEye I fine-tuned GPT-2 to study synthetic disinformation; at Google Cloud Security (Mandiant) I built ML models for threat intelligence that improved detection rates by ~30% and cut false positives by ~20%. Today I focus on the security of LLMs and autonomous agents — a space where the attack surface has expanded fast and the bar to compromise these systems has dropped. I am especially interested in the blind spots of generative AI and in building robust, secure AI ecosystems.

Download my resumé.