The aim of this master's thesis is to speed up the data preparation process for a complete vehicle CFD simulation. Data preparation is the bottleneck in current industrial CFD processes. Preparing a complete vehicle geometry for external or internal flow investigation keeps an engineer busy for several days. To speed up data preparation, the thesis will investigate whether machine learning models are a feasible approach to automate some portions of this process. In particular, the thesis will focus on how to adapt existing machine learning models for point clouds and meshes to process dense and highly detailed 3D representations of vehicles composed of millions of triangles. These models will be tasked to segment the different parts of the vehicle from a large, predefined set of classes, which will pose additional challenges such as overfitting and generalization.