In this study, we engineer a variant of the flavin-dependent halogenase RebH that catalyzes site- and atroposelective halogenation of 3-aryl-4(3H)-quinazolinones via kinetic or dynamic kinetic resolution. The required directed evolution uses a combination of random and site-saturation mutagenesis, substrate walking using two probe substrates, and a two-tiered screening approach involving analysis of variant conversion and then enantioselectivity of improved variants. The resulting variant, 3-T, provides >99:1 e.r. for the (M)-atropisomer of the major brominated product, 25-fold improved conversion, and 91-fold improved site-selectivity relative to the parent enzyme on the probe substrate used in the final rounds of evolution. This high activity and selectivity translates well to several additional substrates with varied steric and electronic properties. Computational modeling and docking simulations are used to rationalize the effects of key mutations on substrate scope and site- and atroposelectivity. Given the range of substrates that have been used for atroposelective synthesis via electrophilic halogenation, these results suggest that FDHs could find many additional applications for atroposelective catalysis. More broadly, this study highlights how RebH can be engineered to accept structurally diverse substrates that enable its use for enantioselective catalysis.