Multidimensional modeling of in-cylinder processes has traditionally relied upon comparison with experimentally determined gross quantities, such as swirl ratio or valve discharge coefficient. Recent experimental studies have focused on accurate in-cylinder measurement of quantities such as velocity fields, species concentration distributions and distributions or turbulent kinetic energy. Since the most important engine design parameters, including filling efficiency, flame stability and pollutant formation depend on the local flow field, the ability to accurately predict these details is a key requirement for successful application of computational fluid dynamics techniques to engine design. One key barrier to accurately resolving local flow details has been the difficulty involved with creating a computational mesh which provides reasonable geometric fidelity and significant resolution of gradients of flow quantities without being so large that it is impractical for use with commonly available engineering computing resources. In this work, a procedure is outlined for producing a computational mesh for a multi-valve intake port and cylinder geometry. Techniques by which the mesh may be subsequently refined based upon the solution itself are also demonstrated. The relative success of the various computational approaches is evaluated through comparison with experimentally obtained local velocity and turbulence distribution data.