This project explores new aspects of supergravity theories in relation with string theory. It consists of two main parts, each with specific objectives. The first part studies novel structures in the AdS/CFT correspondence, focusing on supergravity solutions dual to supersymmetric field theories on curved manifolds. We will construct new supersymmetric, asymptotically locally (Euclidean) Anti de Sitter supergravity solutions, and compute the holographic partition function and other relevant observables of the field theory living on the boundary; these will be compared with a field theory path-integral computation, recently made feasible by the technique of supersymmetric localization. The second part of the project studies supersymmetry breaking in string theory compactifications with fluxes, with the purpose of finding a fully satisfactory de Sitter solution to string theory. This ambitious goal, crucially relating string theory and cosmology, will be pursued using a combination of supergravity techniques and flux compactifications methods. First we will analyse within four-dimensional extended supergravity under what conditions supersymmetry can be broken generating a metastable de Sitter vacuum; then we will study the uplift of these models to string theory.