Abstract: With the advancement of computational resources and methodologies, computational materials science has significantly reinforced experimental efforts and accelerated materials research and development. However, a significant disparity exists between experimental observations and theoretical calculations, primarily because of the structural simplifications often employed in computational models to enhance feasibility. Bridging this gap is challenging, especially when dealing with large, complex systems such as nanoparticles and interfaces. This requires solutions that extend computational simulations to emulate actual systems. In this study, we propose a method that utilizes the moment tensor potential (MTP) combined with active learning techniques for highly reliable and large-scale simulations of alloy nanoparticle catalysts and reactive dynamics at electrode interfaces.