The segregation of solute atoms at grain boundaries (GBs) can profoundly impact the structural properties of metallic alloys, and induce effects that range from strengthening to embrittlement. And, as such, the control of solute segregation is emerging as an alloy design tool, uses of which include the stabilization of nanocrystalline alloys. To date, the standard approach to predict the extent of solute segregation at GBs uses a simplified representation that treats the GB network as a single entity, and thus, uses a single “average” segregation energy to characterize solute GB segregation in an alloy. This simplification, however, fails to capture the highly disordered and anisotropic nature of GBs in polycrystals, which results in a spectrum of solute segregation tendencies (energies). In this thesis, we aim to address and remove this simplification; the thesis has five major contributions. First, we elucidate computationally the nature of this spectrum for an Mg solute in an Al polycrystal; the distribution is found to be captured accurately with a skew-normal function. Second, we outline a thermodynamic segregation isotherm that incorporates this spectrum, and employ it to study the effect of such a spectrum on predictions of the equilibrium GB segregation state. Third, we develop a machine learning framework that can accurately predict the segregation tendency of solute atoms at GB sites in polycrystals, based solely on the undecorated (pre-segregation) local atomic environment of such sites. We proceed to use the learning framework to scan across the alloy space, and build an extensive database of segregation energy spectra for more than 250 metal-based binary alloys. Fourth, we outline more formally correct thermodynamic criteria to screen for thermodynamic stability of polycrystalline structures, accounting for the spectral nature of GBs. And, we proceed to apply the developed criteria to screen over 200 alloy combinations. Among its benefits, this spectral approach enables strict enforcement of the third law of thermodynamics, where an average segregation energy does not. Fifth, we take the first step to extend the developed framework to handle solute segregation beyond the dilute limit, by outlining a thermodynamic segregation isotherm that accounts for both the spectrality of grain boundary sites, and solute-solute interactions; we also develop a computational framework to extract, and delineate both effects. Finally, we hope that the developed spectral thermodynamic framework, machine learning models, and solute segregation database in this thesis would help unlock the full potential of GB segregation as an alloy design tool, and enable the design of microstructures that maximize the useful impacts of segregation.Ph.D.