In many applications metallic structural materials face hydrogen-containing environment and at some point the hydrogen enters the metal leading to mechanical properties deterioration and eventually to rupture. The mechanisms of hydrogen embrittlement have been widely studied. Yet, a general, predictive and quantitative model of these phenomena is still missing. This thesis focuses on hydrogen segregation at grain boundaries which is one of the mechanisms identified in hydrogen embrittlement. We aim at modeling the kinetics of the segregation process starting down from the atomic scale. In order to do this, we need to find the equilibrium structures of grain boundaries, identify the segregation sites for each grain boundary and then quantify how each grain boundary affects the diffusion coefficient of hydrogen. All this data will then be fed to a finite element model whose purpose is to compute hydrogen distribution in a polycristalline sample as a function of time, accounting for the specific properties of each grain boundary. These results will be compared with hydrogen permeation experiments which give access to an effective diffusion coefficient, as well as measures localized around a single grain boundary (PANI and SKPFM methods).