The design of an electrified powertrain, or electric drive for an electric vehicle aims at achieving performance targets in multiple physics domains, including electromagnetics, structures, thermal fluids, and multibody dynamics. As numerical analyses in these areas based on discretized 2- or 3-dimensional models for sufficient accuracy are often costly for a large number of simulations, it is natural to adopt suitable models and simulation methodologies differing in their fidelity and scope at different design stages even for a single physics domain. As many of the performance targets are inter-related across different physics domains, for example electric machine efficiency, maximum stator temperature, and the eigenmodes of the entire drive unit, traditional iterative, component-centric design optimization can easily overlook the maximum design potentials. In addition, at the later design stages it becomes very difficult to assess alternative optimums and what-if scenarios retrospectively. However, extending the design space in earlier design stages is challenging in part due to their higher design uncertainty compared to the final prototypes and difficulty in ensuring consistent data between models. Ideally the choice of model fidelity in numerical simulations for design exploration or validation should be flexible enough to allow for desired balancing between cost and accuracy. The cost is not only for the execution of numerical simulations but also for creating and managing the geometry model with an appropriate level of details along with simulation model data. This work demonstrates a unified approach to modeling and simulation workflow management addressing different levels of modeling fidelity efficiently. The fidelity variants are based on the level of geometric details, degree of parametric freedom, and accuracy of simulation methodology. A parametric, modular electric drive model is associated with discretized simulation models and the analysis of system’s noise and vibration behavior. In addition, the multi-fidelity results and costs are compared from the perspective of typical parametric design studies.