During the design phase of a vehicle, particularly regarding its dynamic behavior, engineers face a wide range of variables and distinct configurations. Intuitively, defining these parameters to meet pre-established project targets can be a challenging task. Moreover, the constant need to obtain valuable information from vehicle telemetry further complicates the process. Simultaneously, the growing use of Machine Learning (ML) algorithms, frequently employed in such cases, is highlighted. Those methods rely on identifying patterns within a pre-existing database, and, regardless of the specific technique applied, if the input data lacks meaningful relationships, Artificial Intelligence (AI) approaches are unlikely to produce satisfactory results. Considering the often utilized nature of these tools as a notable change across various fields, especially if combined with Data Science, developing virtual vehicle models with the aid of those techniques becomes an interesting solution for automotive engineering (AE). In this context, this paper proposes a methodology for creating virtual vehicle models, derived from a baseline sports utility vehicle (SUV), using VI-CarRealTime (VICRT). Four distinct models were developed, varying in parameters such as brake friction, weight, and spring preload. The results demonstrate that the vehicle’s dynamic behavior was influenced as expected by the imposed perturbations. To validate these findings, widely recognized maneuvers from the literature were employed, sweep steer, step steer, J-turn, and fishhook, e.g.