The design process of power switches usually has time constraints, relies on computationally extensive simulation models and is performed for each application separately. A well-defined product is more likely to lead to a more efficient design and reduce the time-to-market by checking for application compatibility and parameter optimization in advance. The product definition space encompasses the multidimensional parameter space where specifications, constraints, and trade-offs of a product's design and functionality are defined and optimized. This work aims at leveraging machine learning surrogate models to discover this space in minutes rather than weeks, provides at least 99% coverage of solution spaces while having < 2% Pareto front error.