We predict, from computer modeling and simulation in partnership with experiment, a general strategy for synthesizing spherical oxide nanocrystals via crystallization from melt. In particular we "simulate synthesis" to generate full atomistic models of undoped and Ti-doped CeO2 nanoparticles, nanorods, and nanoporous framework architectures. Our simulations demonstrate, in quantitative agreement with experiment [Science 2006, 312, 1504], that Ti (dopant) ions change the shape of CeO2 nanocrystals from polyhedral to spherical. We rationalize this morphological change by elucidating, at the atomistic level, the mechanism underpinning its synthesis. In particular, CeO2 nanocrystals can be synthesized via crystallization from melt: as a molten (undoped) CeO2 nanoparticle is cooled, nucleating seeds spontaneously evolve at the surface and express energetically stable {111} facets to minimize the energy. As crystallization proceeds, the {111} facets grow, thus facilitating a polyhedral shape. Conversely, when doped with Ti, a (predominantly) TiO2 shell encapsulates the inner CeO2 core. This shell inhibits the evolution of nucleating seeds at the surface thus rendering it amorphous during cooling. Accordingly, crystallization is forced to proceed via the evolution of a nucleating seed in the bulk CeO2 region of the nanoparticle, and as this seed grows, it remains surrounded by amorphous ions, which "wrap" around the core so that the energies for high-index facets are drastically reduced; these amorphous ions adopt a spherical shape to minimize the surface energy. Crystallization emanates radially from the nucleating seed, and because it is encapsulated by an amorphous shell, the crystallization front is not compelled to express energetically favorable surfaces. Accordingly, after the nanoparticle has crystallized it retains this spherical shape. A typical animation showing the crystallization (with atomistic detail) is available as Supporting Information. From this data we predict that spherical oxide nanocrystals can be synthesized via crystallization from melt in general by suppressing nucleating seed evolution at the surface thus forcing the nucleating seed to spontaneously evolve in the bulk. Nanospheres can, similar to zeolitic classifications, constitute Secondary Building Units (SBUs) and can aggregate to form nanorods and nanoporous framework architectures. Here we have attempted to simulate this process to generate models for CeO2 and Ti-doped CeO2 nanorods and framework architectures. In particular, we predict that Ti doping will "smooth" the surfaces: hexagonal prism shaped CeO2 nanorods with {111} and {100} surfaces become cylindrical, and framework architectures change from facetted pores and channels with well-defined {111} and {100} surfaces to "smooth" pores and channels (expressing both concave and convex curvatures). Such structures are difficult to characterize using, for example, Miller indices; rather we suggest that these new structural materials may be better described using minimal surfaces.