Glasses are widely used for the immobilization of nuclear waste. However, the effect of compositions on the structure and properties of glass remains poorly understood. Here, we employ the latest advancements in the field of artificial intelligence, machine learning, and atomistic simulations to enable an optimal design of the nuclear waste matrix. Specifically, we employ natural language processing to extract a large corpus of literature on glasses used for nuclear waste immobilization. Further, by analyzing these manuscripts, we extract datasets on composition, structure (such as Qn distribution), and properties, such as dissolution rate. Further, we employ machine-learning approaches to develop a predictive model, which enables the design of glass composition with targetted dissolution rate. Finally, we employ atomistic simulations to understand the role of medium-range order in governing the dissolution behavior of glasses.