Composites based on conductive nanowires embedded in elastomers are popular in a wide range of stretchable electronics applications where the requirements are either a stable or a highly increasing electrical resistance upon strain. Despite the widespread use of such composites, their production is not based in solid theoretical grounds but rather in empirical observations. The lack of such a framework is due to limitations in the methods for studying nanowire meshes, in particular the lack of knowledge on the spatial distribution of the nanowires and the change of their position under strain. This hurdle is overcome by collecting 3D reconstructed X-ray tomographies of silver nanowires embedded in polydimethylsiloxane (PDMS) under variable deformations and the missing structural information of the nanomaterial is obtained by unsupervised artificial intelligence image analysis. This allowed to reveal the precise assembly mechanisms of nanowire systems and derive a precise analytical formula for the piezoresistive response of the composite and finally to simulate the behavior of arbitrary samples in-silico.