In advanced high-performance applications, such as HEMTs and SPADs, device operation is often governed by high-field transport regimes, where semi-classical Drift-Diffusion models often fail to capture essential non-equilibrium phenomena. This thesis details the development and implementation of a full-band Monte Carlo (FBMC) simulation framework, centered on the integration of fundamental scattering physics and built upon a data-driven approach, utilizing structured datasets to resolve physical attributes such as energy bands and scattering rates. To address the significant memory constraints associated with storing these datasets, this work further introduces an AI-based implementation, leveraging machine learning techniques to interpolate and compress these datasets, improving memory usage.