Nitrate (NO3–) is a ubiquitous contaminant in groundwater that causes serious public health issues around the world. Though various strategies are able to reduce NO3– to nitrite (NO2–), a rational catalyst design strategy for NO2– removal has not been found, in part because of the complicated reaction network of nitrate chemistry. In this study, we show, through catalytic modeling with density functional theory (DFT) calculations, that the performance of mono- and bimetallic surfaces for nitrite reduction can be rapidly screened using N, N2, and NH3 binding energies as reactivity descriptors. With a number of active surface atomic ensembles identified for nitrite reduction, we have designed a series of “metal-on-metal” bimetallics with optimized surface reactivity and a maximum number of active sites. Choosing Pd-on-Au nanoparticles (NPs) as candidate catalysts, both theory and experiment find that a thin monolayer of Pd-on-Au NPs (size: ∼4 nm) leads to high nitrite reduction performance, outperforming pure Pd NPs and the other Pd surface compositions considered. Experiments show that this thin layer of Pd-on-Au has a relatively high selectivity for N2 formation, compared to pure Pd NPs. More importantly, our study shows that a simple model, based upon DFT-calculated thermodynamic energies, can facilitate catalysts design relevant to environmental issues.