Andy Zhang is a researcher at Stanford University focusing on developing and benchmarking cybersecurity agents. He previously led Cybench, a framework to evaluate cybersecurity capabilities and risk of language model agents and investigated the train-test overlap transparency of language models.