We present a multiscale approach for the identification of novel zeolites as well as for the optimization of simulated moving bed (SMB) chromatographic processes for p-xylene separation. Our hierarchical in silico approach pertains to detailed modeling, simulation, and optimization at molecular and process scales. At the molecular scale, it includes geometric-level pore topology characterization, minimum-energy pathway based selectivity prediction, and shape-selectivity based zeolite ranking for separation of xylenes. At the process scale, first principles model-based optimization of industrial-scale SMB process with 24 columns is performed for the selection of top zeolites and optimal process conditions. We develop a generalized mixed-integer nonlinear optimization (MINLP) model which considers both process design and material selection. We also allow for different zeolites to be selected in different columns. The application of the overall framework results in the discovery of several novel zeolites for the most profitable p-xylene separation using SMB chromatography. The top-ranked shape-selective zeolites are OBW, MEL, MWW, MTT, OWE, FER, SZR, IMF, TON and LAU, which have not been considered in the previous literature for p-xylene separation. To that end, we suggest new applications for these zeolites. Our results indicate that significant increase in profit can be achieved by replacing the current adsorbents by MWW and MEL zeolites. Remarkably, a new SMB process with multiples zeolites introduced in different columns