Mingxi’s research focuses on the algorithmic design and economic analysis of multi-agent collaborative learning. On the algorithmic side, she develops scalable optimization methods for distributed and large-scale training, with an emphasis on communication-efficiency, heterogeneity robustness, and convergence guarantees. On the economic side, she studies incentive and governance problems that arise in collaborative learning, particularly in settings enabled by privacy-enhancing technologies such as federated learning, including how to evaluate and allocate data/agent contributions to a shared model under business context. Her work is motivated by business applications where learning and strategic behavior interact, including online finance and auction platforms.