Recent advances in artificial intelligence and machine learning have led to the development of many tools for implementing and using neural networks. One field of interest is physics simulation, where systems governed by Partial Differential Equations (PDE) and Ordinary Differential Equations (ODE) are solved and approximated by conventional numerical solutions. A modern approach to solving such systems includes using machine-learned Neural Networks (NN) trained on recorded data and known governing equations of a system. The resulting NN is a Physics-Informed Neural Networks (PINN) and can be used as an efficient surrogate solver for the original system. One such software is the open-source Python library Nvidia Modulus (recently renamed Nvidia PhysicsNeMo), enabling users to implement, execute, and train such PINNs. This thesis investigates whether and how a modulus model could be implemented and coupled through the physics simulation coupling library preCICE to solve a simple partitioned heat problem. The result provides the groundwork for future implementations where parts of a multi-physics system do not have a reasonable or efficient known conventional solution and warrant such a surrogate solution.