The unprecedented recent success of artificial intelligence and machine learning has largely been driven by centralized models trained on massive, aggregated datasets. However, as data collection and processing increasingly occur at the edge—within complex, interconnected environments such as smart grids, autonomous computer clusters, and the Internet of Things (IoT)—significant new challenges emerge. Centralized paradigms often struggle with high communication costs, latency bottlenecks, and stringent data privacy regulations, making the transition toward efficient, decentralized learning a critical frontier for the next generation of AI. This project, funded by the Swedish Research Council, aims to advance the intersection of federated learning, reinforcement learning, and multi-agent systems by developing novel algorithms and rigorous theoretical frameworks for distributed AI. To achieve this, the research will seamlessly integrate techniques from mathematical optimization, deep learning, and statistical learning. This interdisciplinary approach is designed to address key challenges inherent in learning from distributed, privacy-sensitive, and dynamically evolving data. Specifically, the project will tackle the complexities of decision-making in non-stationary environments, handling heterogeneous data distributions across edge devices, and ensuring system robustness against uncertainties and potential adversarial disruptions. Furthermore, the project bridges the gap between fundamental mathematical theory and practical deployment. Focusing heavily on both algorithmic development and real-world applications, the research draws upon methodologies from optimization, advanced simulation, control theory, and distributed computing. By applying these decentralized learning frameworks to critical physical systems, the project aims to yield intelligent agents that are not only highly scalable and performant but also interpretable, secure, and aligned with modern data privacy standards. Ultimately, the project is part of a vibrant academic community characterized by a strong international network. It offers an exceptional environment for interdisciplinary collaboration, continuous scientific exchange, and joint initiatives bridging academia and industry, ensuring that theoretical breakthroughs in distributed AI translate into tangible, real-world impact.