While molecular materials are increasingly finding utility in energy applications, changes in molecular packing between different polymorphs can result in large variations in physical and chemical properties. Thus, one cannot optimize functionality at the single-molecule level; the crystal packing must be considered to optimize properties of a molecular material for targeted applications. Dispersion-corrected density-functional theory (DFT) has shown great promise in resolving the small energy differences between polymorphs required for successful first-principles crystal-structure prediction (CSP). However, the computational cost makes full plane-wave/pseudopotential DFT calculations impractical to apply to the vast number of candidate structures considered in a CSP search. Composite approaches, combining low-level geometry relaxation with high-level energy calculations, can greatly increase the efficiency of a CSP protocol. Additionally, numerical atom-centered orbitals (NAOs) provide significant time and memory savings over plane-wave basis sets and enable routine use of hybrid functionals. In this talk, progress towards CSP using DFT and the exchange-hole dipole moment (XDM) dispersion correction will be highlighted, as will recent work on crystal-property prediction through modeling polymorph and co-crystal dependent photoluminescence.