2D materials present a compelling alternative to other solid lubricants thanks to their unique properties: high specific surface area, in-plane strength, weak interlayer interactions, and surface chemical stability. These properties translate to remarkably low friction and high wear resistance, ultimately leading to improved energy efficiency, which has both environmental and economic benefits. Heterostructures are an avenue for consistently achieving low friction thanks to forced incommensurability. Matching heterointerfaces is crucial for creating novel material properties and functionalities. This work explores the systematic use of Machine Learning Interatomic Potentials (MLIPs), specifically MACE, to perform a high-throughput Non-Equilibrium Molecular Dynamics (NEMD) review of friction in 2D material heterostructures using a large range of materials from 2Dmatpedia, C2DB, and MC2D. The python-based program FrictionSim2D, previously created, is used to conduct the high-throughput workflow. This workflow automates the generation of simulation cells, interatomic potential parameters, and input files for NEMD simulations to conduct Prandtl-Tomlinson-like AFM tip - sheet -substrate simulations and sheet on sheet simulations in LAMMPS. The extracted tribological data will be used to train ML models, such as regressive random forests, LSTMS, and LLMs, to predict the performance of new materials and heterostructures. Our goal is to establish correlations between specific material properties and atomic friction mechanisms, gaining deeper insights into the underlying causes of atomic friction. We anticipate that this project will help accelerate the design, prototyping, and experimental validation of 2D materials that demonstrate robust superlubricity, paving the way for their widespread adoption in various applications. Furthermore, the established workflow will greatly help with the parameterization and improvement of MLIPs for both frictional applications and use on 2D materials. In this presentation, I will delve into the details of our framework including MLIP implementation and parameterization, and demonstrate its ability to interpret NEMD-derived tribological data to gain important insights on novel 2D material heterostructures based on their intrinsic properties.