This $275,000 project grant awarded by the National Science Foundation (NSF) Engineering program (CFDA 47.041) is for collaborative research on new algorithms and theory for weakly coupled Markov decision processes (WCMDPs). The research aims to develop theoretical foundations and innovative algorithms to "decouple" and "reassemble" large WCMDPs, which are used to model complex systems across fields like job scheduling, resource allocation, and supply chain management. The research will draw on new "one-to-many" approaches as well as classical techniques from large stochastic systems and reinforcement learning. The algorithms and theory developed will be evaluated through simulated problems and a real-world resource management problem using data from Google's datacenters. This research is expected to advance the traditional algorithms and theory for WCMDPs and large-scale Markov decision processes more broadly. The project will also include curriculum development, mentoring programs, and diversity initiatives to recruit students into this area of research.