Optimal design strives to identify the best-performing devices within a given metric, a problem common to all engineering branches. Considering radio frequencies, the challenge might be to find an antenna of a given bandwidth, matching, gain, and electrical size. Unfortunately, optimal design is strictly unfeasible in polynomial time. Many approaches have been adopted over the years to solve such problems in a satisfactory manner, yet our requirements for modern devices push contemporary techniques to their limits eventually rendering them obsolete. In this lecture, the optimal design task will be formulated via the so-called exact reanalysis technique. Altogether, the novel framework deals with non-linear problems by efficiently collecting large data sets for further post-processing based on AI-assisted systems. Its implementation into the commercialized toolbox AToM is demonstrated and accompanied by several showcases involving the inverse design of electrically small antennas, planar lenses, and electromagnetic cloaks. The lecture concludes with a discussion of further advancements in the design of analog wave computers or reconfigurable intelligent surfaces, future technologies that will, undoubtedly, disrupt development in electromagnetics and radio electronics.