Size enlargement control and modeling in fluidized beds are crucial in the pharmaceutical and food industries but remain underdeveloped for technical catalyst formulation and shaping. This work uses different modeling approaches to understand aggregation kinetics: single- and two-pathway population balance equation (PBE) modeling and machine learning. These models are trained on a large dataset of experimental results from a bottom spray-fluidized bed, using realistic technical catalyst conditions and ingredients: ZSM-5 zeolite, bentonite, and alumina. Our optimized model is based on a two-pathway PBE with two distinct collision efficiencies for early- and late-stage growth dynamics across nucleation, seed formation, seed aggregation, and layered growth. With this model, we discuss the granulation and agglomeration dynamics of realistic technical catalysts and study the controlled shaping of several case studies with tailored morphologies (50, 100, and 200 μm pellets) under optimized conditions (i.e., maximum yield within the desired particle range) as validation.