About my story

My research focuses on AI safety alignment, LLM interpretability, and Human-centered AI, particularly persona, emotion, and social bias in LLMs. I study how LLMs represent emotions and social context, adopt perspectives, express empathy, and respond to persona prompts. Building on my interpretability research, I aim to understand how post-training shapes the internal mechanisms underlying safe and unsafe behavior in LLMs.

Before beginning my graduate studies, I worked as a web frontend engineer.