Over the last three decades, nanocrystalline alloys (polycrystals with grain sizes of less than 100 nm) have been shown to exhibit superior material properties, such as enhanced specific strength, hardness, wear resistance, radiation resistance, and magnetic properties. However, such structures are inherently thermodynamically unstable; a nanocrystalline configuration comes with a large volume fraction of high-energy bearing defects that introduce a large excess of energy in the structure. The key route to overcome this limitation and thermodynamically stabilize nanocrystalline metals against grain growth is through intentional alloying for grain boundary segregation. To date, the standard approach to designing and screening for nanocrystalline stability uses a highly simplified model in which grain boundary networks are treated as a “single” entity, and the tendency of solute atoms to segregate at those boundaries is quantified by an “averaged” value. This simplification, however, ignores the fact that grain boundaries in polycrystals have a vast range of local atomic environments that can attract or repel solute atoms to different degrees. In this talk, I will review our recent efforts to tackle this simplification by developing thermodynamic, computational, and data science frameworks to (i) thoroughly understand the phenomenon of grain boundary segregation at the atomistic scale, (ii) develop comprehensive segregation databases for hundreds of substitutional alloys, and (iii) leverage that knowledge and data into developing rigorous design and screening criteria for nanocrystalline alloys that take into account the spectrality of the grain boundary network.