Ao Qu is currently pursuing his Ph.D. at the MIT Institute for Data, Systems, and Society (IDSS). He received his bachelor’s degree from Vanderbilt University, where he double-majored in Mathematics and Computer Science and graduated with highest honors. His research focuses on self-evolving AI systems, with an emphasis on enabling AI systems to continually learn from interaction, consolidate experience into reusable knowledge, and progressively improve their reasoning and decision-making capabilities.

His research spans agent planning, agent memory, experience-based learning, and self-evolving agents. His work includes the first framework to optimize agent working memory through reinforcement learning, the largest large language model benchmark for operations research to date, and multi-agent systems that have achieved state-of-the-art results across multiple open-ended discovery tasks. He has published more than 15 papers at CCF-A conferences and has received Best Paper Awards at leading academic conferences and workshops, including NeurIPS, KDD, IJCAI, and INFORMS.

He previously worked as a Research Scientist Intern at ByteDance Seed, where he was selected for the TopSeed program and contributed to the development of large-scale continual learning systems.

He is currently building reef, an infrastructure for self-evolving agents that supports continual updates to both model parameters and agent harnesses. The platform is designed to enable real-world applications across engineering optimization, scientific discovery, and continual learning for specialized models.