Liquid–liquid phase separation is a fundamental thermodynamic process that governs structure and function in biological systems, chemical separations, and multiphase materials. Within this broad field, computational studies play a critical role in advancing understanding of this phenomenon by enabling investigation of phase behavior across multiple length scales. In this thesis, liquid–liquid phase separation is explored across scales ranging from atomistic simulations to phase-field models. At the atomistic level, concepts from phase-field theory are used to derive a bias potential that enables direct control of liquid–liquid phase separation in molecular simulations. Within this enhanced sampling framework, macroscopic thermodynamic quantities such as excess free energy, interaction parameters, and activity coefficients are obtained, as demonstrated using binary Lennard–Jones mixtures as model systems. Furthermore, biasing liquid–liquid phase separation under different environmental conditions provides a pathway to relate free energies of phase separation to the stimuli response of Lennard–Jones mixtures and aqueous monomer solutions. Beyond the atomistic scale, liquid–liquid phase separation is examined through simulations of bicontinuous morphologies to generate machine learning–ready datasets for autonomous materials fabrication. In addition, phase-field modeling is applied to examine equilibrium partitioning in ternary liquid mixtures and their dynamics of liquid–liquid phase separation arising from solvent transfer methods. Altogether, this thesis develops computational strategies that extend from atomistic simulations to phase-field modeling, establishing a foundation for linking molecular-level features of liquid–liquid phase separation with macroscopic thermodynamic descriptions. These tools open opportunities to design, optimize, and control multiphase processes and materials across diverse applications.