This paper is focused on the development and application of a CFD methodology that can be applied to predict the fuel-air mixing process in stratified charge, sparkignition engines. The Eulerian-Lagrangian approach was used to model the spray evolution together with a liquid film model that properly takes into account its effects on the fuel-air mixing process into account. However, numerical simulation of stratified combustion in SI engines is a very challenging task for CFD modeling, due to the complex interaction of different physical phenomena involving turbulent, reacting and multiphase flows evolving inside a moving geometry. Hence, for a proper assessment of the different sub-models involved a detailed set of experimental optical data is required.To this end, a large experimental database was built by the authors. In particular, the spray morphology was characterized in detail inside a constant volume vessel, where images were acquired by a CCD camera and then post-processed to evaluate the spray penetration and cone-angles. Furthermore, experiments were carried out in an optically accessible combustion chamber reproducing a real 4-stroke, 4-cylinder, high performance GDI engine. The cylinder head was instrumented by using an endoscopic system coupled to high spatial and temporal resolution cameras in order to allow the visualization of the fuel injection and the combustion process. The complete set of spray models was tuned with experiments carried out at constant-volume conditions, then full-cycle simulations were performed for the optical engine. Four different operating points were simulated accounting for different injection pressures and charge stratification levels. Validation was carried out by comparing computed and experimental data of spray and liquid film evolutions. To further verify the computed results, computed equivalence ratio distributions at spark-timing were correlated with optical images of flame propagation.