Advanced artificial intelligence (AI) and machine learning (ML) models are rapidly transforming meteorology, particularly in high-impact weather prediction and analysis. This session explores cutting-edge AI/ML methodologies, including foundation models, vision transformers, physics-informed neural networks, and diffusion-based generative models, applied to forecasting severe weather events such as hurricanes, tornadoes, extreme rainfall, blizzards, and wildfires. We discuss how modern architectures—including transformer-based models and graph neural networks—extract intricate spatiotemporal patterns and nonlinear relationships that traditional models may overlook, improving physical understanding and predictive capabilities. The session highlights innovative approaches leveraging self-supervised learning, multi-modal fusion, and physics-guided regularization to assimilate heterogeneous data streams (satellite, radar, surface observations), better representing fine-grained atmospheric processes. Topics include foundation model adaptation, hybrid physics-ML frameworks, vision transformer-based quantitative precipitation estimation, deep learning emulators for numerical model acceleration, and AI-enhanced regional climate projection.