This project focuses on the development and training of machine learning models to predict electromagnetic behavior of antenna structures using synthetic simulation data. Input data consists of binary-encoded antenna layouts represented as 17×17 grids, while the output consists of complex S-parameters (4×4×36), generated through full-wave electromagnetic simulations. These datasets are stored in MATLAB .mat and .csv formats and will be used to train deep neural networks capable of learning the nonlinear relationship between antenna geometry and its frequency response. The primary goal is to enable fast and accurate prediction of antenna performance without repeated simulations, thereby accelerating the design and optimization process. The project will rely on extensive data preprocessing, model training using TensorFlow, and storage of large intermediate datasets. An upgrade to 200 GiB of storage is required to accommodate high-volume data generation, processing, and archiving. This research contributes to the field of computational electromagnetics and data-driven RF design, with potential applications in wireless communication systems and automated antenna synthesis.