Abstract: I will discuss recent efforts to overcome challenges in the prediction of electronic processes in soft materials through the development of quantum chemical models that operate at coarse-grained resolutions. We motivate the origins of our computational approach, denoted electronic coarse-graining (ECG), discuss its relationship to existing molecular modeling frameworks, and describe recent successes for soft materials. Specifically, we will show how the use of Deep Kernel Learning enables a quantitative connection to underlying all-atom models, exactly producing the all-atom quantum chemical observables, with fluctuations, at the coarse-grained resolution. We will then highlight the use of a novel message passing algorithm, denoted continuously gated message passing (CGMP), that enables the one-shot optimization of coarse-grained mappings for arbitrary quantum chemical prediction tasks. We will show how CGMP can be used with ECG methods to provide scalable calculations of electronic observables using only coarse-grained representations, enabling the first-principles treatment of electronic transport in liquid crystalline molecular semiconductors.