Modern computers run applications in "containers", software units that contain the application code and all the additional software that the code requires to run. To manage these containerized applications, sophisticated platforms such as Kubernetes have become crucial for ensuring scalability, reliability, and cost efficiency. However, tuning these platforms for optimal performance is a highly complex task, given their large configuration spaces and subtle parameter interactions. This project addresses these challenges by developing Cosimo, a co-simulation-based optimization framework that integrates real-world orchestration platforms with simulated containerized applications. By enabling low-overhead, high-fidelity experiments without the need for costly physical testbeds, Cosimo empowers researchers and practitioners to optimize orchestration strategies in a scalable, efficient, and cost-effective way. The broader impacts of the project include advancing cloud computing practices, improving resource efficiency in both scientific and industrial cloud environments, and contributing to workforce development through integration into graduate education, open-source dissemination, and public outreach. The project introduces Cosimo as a transformative tool for understanding and improving container orchestration at scale. It involves three main research thrusts: (1) developing a co-simulation framework that models critical orchestration features such as autoscaling, load balanc