My research focuses on developing computational methods for optimization and learning in complex decision systems. In particular, I study hierarchical and distributed stochastic optimization, where decisions are made across multiple levels or by multiple interacting agents under uncertainty and limited information. I also work on federated and personalized learning, where multiple clients collaboratively train models while retaining local data. These methodological developments are motivated by applications in multi-agent systems, transportation networks, logistics, resource allocation, and over-parameterized machine learning.