Before being released to the market, all vehicle prototypes are validated in terms of their crashworthiness. Meeting the safety standards, while same time avoiding compromises in other essential design parameters, requires a very meticulous engineering simulation approach during product design. These processes become even more complicated, and time-consuming, with electric vehicles, such as Lithium-ion battery-powered cars. In many cases to accomplish the safety aims during product design, while also meeting time limitations and deadlines, sophisticated simulation tools need to be employed. Such tools are those that enable optimization studies and that take advantage of Machine Learning capabilities. In this study, an optimization and a quick verification of an electric vehicle side frame design were performed with the aid of an Optimization tool and Machine Learning methods. Through this Optimization tool several Design Experiments have been created and then, by training a Machine Learning model, also referred to as a “Predictor”, the optimal design parameters were approached for the given objective and constraints. Since during the designing stages, the geometry can be often modified, the proposed approach saves considerable amounts of time, as it avoids repeating the complete ML Optimization process or solving each updated model individually. To accomplish this, Transfer Learning methods were utilized in order to employ the already trained ML predictive model to verify and optimize the updated geometry. In this way, the optimal design for the updated model was calculated, avoiding re-training an updated Predictor by producing new data sets. The use of this already trained Predictor extends also to the field of a quick verification of the newly updated designs. The several design modifications were quickly tested without needing to solve the model again. To further reduce simulation time and modeling effort, a macroscopic battery model was used. This way, the ML Predictor was able to also consider the electromagnetic phenomena related to damaged batteries without increasing the solution time of the side-crash simulation. All in all, using the Machine Learning based Optimization tool, and the Transfer Learning related functionality, an already trained predictive model was able to estimate the optimal design of the vehicle with updated components and verified the updated designs without having to re-run the complete optimization and solution processes.