As modern video games transition toward increasingly dense and interactive environments, traditional Object-Oriented Programming (OOP) patterns in game engines face significant CPU bottlenecks. This thesis investigates the implementation of a high-performance large-scale agent simulation system using Unreal Engine’s Mass Framework, a data-oriented suite leveraging Entity Component System (ECS) architecture. The research aims to achieve a scalable simulation of thousands of agents on mid-range hardware, focusing on several key technical challenges: seamless navigation across non-planar landscapes, dynamic interaction with level objects, and a hybrid movement system capable of transitioning from global path-following to local target acquisition. The implementation utilizes MassEntity for memory-efficient data management and explores the integration of diverse animation states to maintain visual fidelity at scale.