Studying nucleation is crucial for scientific and technological progress, aligning with DOE and Basic Energy Sciences objectives. Nucleation's role in energy and environmental challenges is significant, but its atomic scale understanding and control is challenging. Our proposal combines chemical intuition with AI to comprehensively investigate nucleation in solvated systems. Integrating AI methods like deep variational information bottleneck and Diffusion Probabilistic Models with statistical mechanics approaches, we aim to develop computational methods for enhanced sampling and understanding of both nucleation thermodynamics and kinetics, together with size and supersaturation effects. We will apply these methods to practical systems like Sodium Chloride and colloidal anisotropic systems, surpassing current computational capabilities. Our goal is to provide new insights and solutions for nucleation complexities particularly in solvated systems with static/fluctuating interfaces.