A huge fraction of engineering materials, ranging from steels to ceramics, are composed of crystals. These constituent crystals exhibit a wide variety of defects, which modern materials science exploits to use for tuning of specific material properties. However, for knowledge-driven materials design it is of utmost importance to understand the fundamental mechanisms, how such defects impacts materials. Atomistic modeling combined with first-principles methods is able to provide unique insight into properties such as e.g., local atomic structures. The microstructural complexity of modern materials requires increasingly demanding and sophisticated atomistic models. The standard tools for modelling such complex models however have several shortcomings. Density functional theory (DFT), which gives insight into the electronic structure, is limited to small models. These are small in terms of atoms present and hence fail to describe structurally complex defects. In contrast, molecular dynamics (MD) can handle large models, however falls short in describing the chemical complexity of modern materials such as, e.g. alloys. In the present thesis we have implemented a hybrid approach coupling DFT and MD, to overcome these limitations. We have applied the approach to study interfaces (planar defects) in TiAl intermetallic alloys. Thereby, we study the impact of alloying elements on interfacial mechanical properties. In a second part we tackle the challenge of modeling disordered alloys with finite atomistic models. We provide the theory and a corresponding implementation to generate atomistic models of disordered system. Using the developed tools we create models to study a grain boundary in a (disordered) Ni-base alloy. Besides giving the theoretical prerequisites, we focus on the impact of chemical disorder on solute segregation of alloying elements. Finally, we compare predictions from thermodynamic models between a standard single-species-matrix and the realistic chemically complex matrix.