This research project focuses on the development of a bio-inspired procedural generation technique for electromagnetic components. This work allows for the automated synthesis of structures capable of controlled interactions with incident waves, guiding the growth of components to meet multiple coupling and radiation constraints. The computerization of the means of measurement and production are indeed in several aspects based on the development of dedicated electromagnetic components. Whether they allow these systems to communicate, to interrogate their environment or to direct data flows between their various sub-parts, these components must face the definition of increasingly specific and constrained specifications. They must also respond to the problems of increasing carrier frequency, which facilitates higher data rates and the integration of these solutions. However, such case-by-case optimizations are accompanied by prohibitive engineering costs, justifying that many applications are limited to the use of generic components, which are in essence less well adapted to these environmental and field specifications. This project proposes to develop a technique capable of addressing this issue. This technique will be exploited first to ensure the procedural synthesis of radiating metasurfaces optimized for the realization of multiple polarization constrained point-to-point links and for computational imaging activities. The versatility of the developed generative models will also allow for the procedural synthesis of functionalized materials, suitable for free-form waveguide design and constrained generation of aniostropy properties. This research project will focus on the realization of proofs of concept in K and W bands, demonstrating through the automated design of numerous components the diversity of the potentially impacted domains.