This minisymposium will explore new developments in neural operators for solving partial differential equations (PDEs) in complex geometries. Traditional numerical methods often struggle with the challenges posed by complex geometries, such as handling irregular boundaries, multi-scale features, and intricate topologies. These challenges can lead to increased computational costs, reduced accuracy, and difficulties in generalizing solutions across different domains. Neural operators offer a powerful alternative to traditional PDE solvers, but they inherit similar complexities when dealing with complex domains. This minisymposium aims to bring together experts in machine learning, numerical analysis, and computational physics to discuss novel architectures, training strategies, theoretical insights and practical applications of neural operators, with a focus on addressing the specific difficulties encountered in solving PDEs in complex geometries.