Modern consumer products —­ ranging from coatings and adhesives to personal care formulations — are complex mixtures of polymers, surfactants, solvents, and other active compounds that are carefully tuned for properties such as stability, rheology, and conductivity. Historically, optimizing these “soft matter” formulations has relied on trial-and-error experimentation, which is both time-intensive and costly. Although recent advances in machine learning (ML) and artificial intelligence (AI) show promise for accelerating this process, they are ill-suited for the typically limited experimental datasets, underscoring the need for robust physics-based methods.

In this talk, I will outline a multiscale computational framework that leverages detailed atomistic simulations to inform mesoscale field-theoretic models. We begin by using relative entropy minimization to construct coarse-grained (CG) models from reference atomistic simulations. These CG models are then analytically transformed into equivalent field-theoretic representations, enabling the efficient prediction of solution phase behavior while retaining direct connections to the underlying chemical details. The utility of this approach for guiding experiments is demonstrated on two systems: (1) the phase behavior of acrylic diblock copolymers in dodecane, where in silico insights inform experimental synthesis, and (2) the modeling of hyperbranched polymer synthesis, where an alternative formulation of the field theory allows us to simulate reactive systems. Beyond these applications, I also briefly discuss extensions of this framework to studying formulations with highly branched polymers and solutions under confinement. Overall, this methodology provides a way to accelerate the design of complex, multicomponent formulations with new and sustainable chemistries.