
We study the inverse design of electromagnetic media composed of many small, randomly distributed impedance particles. The underlying direct scattering problem is posed as a complete exterior boundary-value problem and treated within an asymptotic many-body framework, yielding an explicit map between particle-scale parameters and the effective magnetic permeability of the composite. On this basis, the inverse task – constructing a feasible random microstructure that implements a prescribed permeability profile – is formulated as a constrained optimization problem under admissibility conditions and is solved using a genetic algorithm. The explicit asymptotic forward map enables efficient evaluation of candidate configurations, making the inverse design computationally tractable without a full-wave solver inside the optimization loop. Numerical experiments in three-dimensional settings demonstrate that nontrivial spatial permeability profiles can be approximated with mean relative errors below 10% in the tested cases, while respecting the validity conditions of the asymptotic model. Independent benchmark calculations indicate mean validation discrepancies of around 2% within the tested regime. The results show that asymptotic reduced models can provide a practical basis for property-driven design of random particle-loaded electromagnetic media. The framework may be extended to other effective material parameters and related wave-scattering settings when suitable asymptotic forward maps are available.