We have developed a computer algorithm to generate quantitative molecular representations (QMRs) of asphaltenes based on experimental data. First, we generate molecular representations using a Monte Carlo method. For this purpose, we use an extensive set of aromatic and aliphatic building blocks, which are sampled randomly from the corresponding distribution and then linked together using a connection algorithm. The building blocks can be taken from a predefined inventory or generated during run time. Manually prefabricated blocks ensure model flexibility, while automatically generated blocks allow us to build large aromatic sheets. We allow for both archipelago and peri-condensed structures to be generated. Then, we use a nonlinear optimization procedure to select a small subset of molecules that gives the best match with experimental data. These experimental data consist of molecular weight (MW), elemental analysis, and nuclear magnetic resonance (NMR) spectroscopy, including both H-1 and C-13 data. First, we validate the method by testing a number of single model compounds. Then, we use a real asphaltene data set available in the literature. Different values of the MW were used as input parameters. We tested two specific values of the MW in detail, representing the peri-condensed and archipelago structure, respectively: MW = 750 and 4190. For both MWs, we generated 10 sets of 5000 samples each. The samples were then optimized with respect to the experimental objective function. Then, we calculate the value of the objective function as an average over all of the simulation runs. It turns out that the value of the objective function is significantly smaller for MW = 750 than for MW = 4190. This means that the lower MW of 750 provides the best match with the experimental data. As an example, one of the optimized QMR asphaltene structures generated was then used as input in molecular dynamics (MD) simulations to study the formation of nanoaggregates.