At Intelligent Systems Lab, I designed a reinforcement learning method to keep label-free LLM self-training stable against reward hacking. I also developed a long-text classification method based on short, targeted verification questions and evaluation frameworks for LLM-based systems.
My research includes vision-language-action models for a social robot that responds to human emotion in context, and I was first author of an IEEE RO-MAN 2026 paper. Earlier, I conducted OSINT investigations at OPORA Cyber Tech and founded a Military Cyber Defense Unit at IDF Home Front Command.

