Member of Technical Staff, OpenAI
E-mail: csxuandongzhao at gmail
Address: San Francisco, CA, USA
I am a Member of Technical Staff at OpenAI, working on mid- and post-training research to improve the utility and reliability of AI agents. Before that, I was a Postdoctoral Researcher at UC Berkeley as part of the BAIR and RDI, working with Prof. Dawn Song. I earned my PhD in Computer Science from UC Santa Barbara, where I was advised by Prof. Yu-Xiang Wang and Prof. Lei Li. Prior to that, I graduated with a Bachelor's degree in Computer Science from Zhejiang University. I have also interned at Alibaba, Microsoft, and Google.
During my PhD and postdoc, I worked at the intersection of Machine Learning, Natural Language Processing, and AI Safety. My research focused on advancing the capability and reliability of frontier AI models and agents: building increasingly powerful systems through scalable post-training and self-improvement, while ensuring they are safe and aligned with human values.
On the capability side, I developed scalable methods for LLM reasoning and self-improvement, synthetic data pipelines for agentic AI, and rigorous evaluation benchmarks for frontier systems. On the reliability side, I worked across the lifecycle of responsible AI development: robustness to prompt injection and jailbreaks, content provenance via watermarking, and protecting privacy, copyright, and intellectual property in model training and deployment.