Asphaltenes constitute the heaviest aromatic component of crude oil. The myriad of asphaltene molecules falls largely into two conceptual classes: continental-possessing a single polyaromatic core-and archipelago-possessing multiple polyaromatic cores linked by alkyl chains. In this work, we study the influence of molecular architecture upon aggregation behavior and molecular folding of prototypical archipelago asphaltenes using coarse-grained molecular dynamics simulation and nonlinear manifold learning. The mechanistic details of aggregation depend sensitively on the molecular structure. Molecules possessing three polyaromatic cores show a higher aggregation propensity than those with two, and linear archipelago architectures more readily form a fractal network than ring topologies, although the resulting aggregates are more susceptible to disruption by chemical dispersants. The Yen-Mullins hierarchy of self-assembled aggregates is attenuated at high asphaltene mass fractions because of the dominance of promiscuous parallel stacking interactions within a percolating network rather than the formation of rodlike nanoaggregates and nanoaggregate clusters. The resulting spanning porous network possesses a fractal dimension of 1.0 on short length scales and 2.0 on long length scales regardless of the archipelago architecture. The incompatibility of the observed assembly behavior with the Yen-Mullins hierarchy lends support that high-molecular weight archipelago architectures do not occur at significant levels in natural crude oils. Low-dimensional free energy surfaces discovered by nonlinear dimensionality reduction reveal a rich diversity of metastable configurations and folding behavior reminiscent of protein folding and inform how intramolecular structures can be modulated by controlling asphaltene mass fraction and dispersant concentration.