In order to assess the effect of turbulence models in prediction of flow structure with adverse pressure gradient, steady state Reynolds-averaged Navier-Stokes (RANS) equations in an annular axisymmetric diffuser are solved. After selection of the best turbulence model, an approach for the shape optimization of annular diffusers is presented. The goal in our optimization process is to maximize diffuser performance and, in this way, pressure recovery by optimizing the geometry. Our methodology is the optimization through wall contouring of a given two-dimensional diffuser length and area ratio. The developed algorithm uses the CFD software: Fluent for the hydrodynamic analysis and employs surrogate modeling and an expected improvement approach to optimization. The non-uniform rational basic splines (NURBS) are used to represent the shape of diffuser wall with two to ten design variables, respectively. In order to manage solution time, the Kriging surrogate model is employed to predict exact answers. The CFD software and the Kriging model have been combined for a fully automated operation using some special control commands on the Matlab platform. In order to seek a balance between local and global search, an adaptive sample criterion is employed. The optimal design exhibits a reasonable performance improvement compared with the reference design.