Mechanical properties of composite materials are often hierarchical emerging from atomic-level structures, chemical and physical interactions, dimensions and distributions of ordered or disordered domains, and dynamics over time scales relevant to the process or system. Identifying correlated structure-property relationships is central to the rational design of advanced structural composites, but is challenging for systems with heterogeneous compositions, poor long-range order, and low particle surface areas. Advanced solid-state nuclear magnetic resonance (NMR) techniques are sensitive to ordered and disordered environments and enable the preferential detection of surface species or interactions, the resolution of distinct local atomic structures in bulk materials, and the measurement of domain sizes ranging from <1nm to micrometer length scales. In contrast to conventional NMR, Dynamic Nuclear Polarization (DNP) involves the manipulation of coupled electron-nuclear spin ensembles to generate a large nuclear magnetization gradient that dramatically enhances NMR signal sensitivity. An advantage of DNP-NMR is that hyperpolarization emanates from paramagnetic centers, enabling surface-enhanced NMR spectroscopy of porous or nonporous solids. Despite the widespread application of this technique, quantitative models of DNP polarization transfer are difficult to implement and rely on coupled spin polarization generation, propagation, and dissipation processes that bridge quantum mechanical and classical phenomena. In Chapter two and three of this manuscript, a constitutive model is derived to quantitatively describe aspects of mesoscopic spin polarization transfer. From dimensional property analysis, spin polarization analogs of the dimensionless Biot number (Bi), Hatta number (Ha), Damköhler number (Da), and Thiele modulus (????) are discovered and their general relevance to spin polarization transport processes demonstrated. Importantly, by analogy to heat and mass transfer film theory, a DNP transfer coefficient, kDNP, with units of m/s is empirically measured and reveals new insights into the transfer of spin hyperpolarization across the electron-nuclear spin interface (i.e., spin-diffusion barrier). In Chapter four and five of this manuscript, combined DNP-enhanced 1H-1H spin diffusion experiments and modelling analyses are used to measure the compositions and dimensions of silicate hydrates which form on tricalcium silicate particles in contact with water. The dimensions and compositions of these surface hydrates are believed to crucially influence the early hydration kinetics of industrially relevant cement-water mixtures, thus by understanding these processes at the atomic-level, new criteria is provided to inform chemical admixture design and to aid in the prediction of hydration rates. Lastly, in Chapter six of this manuscript, advanced two-dimensional 27Al{29Si} heteronuclear correlation experiments are used to monitor hydration processes involving volcanic glasses in cementitious mixtures. These analyses provide new geochemical insights regarding the structure of vitreous pyroclastic minerals and informs the design of modern Roman-inspired pozzolanic concretes with improved longevity and a lower carbon footprint in comparison to traditional cements.