Geometric structures and dualities arise naturally in quantum field theories and string theory. In fact, these tools become very useful when studying strong coupling effects, where standard perturbative techniques can no longer be used. In this thesis we look at several conformal field theories in various dimensions. We first discuss the structure of the nilpotent networks stemming from T-brane deformations in 4D N=1 theories and then go to the stringy origins of 6D superconformal field theories to realize deformations associated with T-branes in terms of simple combinatorial data. We then analyze non-perturbative generalizations of orientifold 3-planes (i.e. S-folds) in order to produce different 4D N=2 theories. Afterwards, we turn our attention towards a few dualities found at strong coupling. For instance, abelian T-duality is known to be a full duality in string theory between type IIA and type IIB. Its nonabelian generalization, Poisson-Lie T-duality, has only been conjectured to be so. We show that Poisson-Lie symmetric sigma-models are at least two-loop renormalizable and their beta-functions are invariant under Poisson-Lie T-duality. Finally, we review recent progress leading to phenomenologically relevant dualities between M-theory on local G_2 spaces and F-theory on locally elliptically fibered Calabi-Yau fourfolds. In particular, we find that the 3D N=1 effective field theory defined by M-theory on a local Spin(7) space unifies the Higgs bundle data associated with 4D N=1 M-theory and F-theory vacua. We finish with some comments on 3D interfaces at strong coupling.