Due to the physical phenomena involved in Powder Bed Fusion (PBF) processes, the surface of the manufactured product is naturally rough, significantly impacting the characteristics of fluid flow and heat transfer over the surface. Several studies have attempted to develop Computational Fluid Dynamics (CFD) models to predict friction factor and Nusselt number in rough channels and despite progress in roughness models, further investigations are required to develop a numerically reliable method in the study of heat transfer over the highly irregular roughness made by PBF. This study developed a high-fidelity numerical model based on the Reynolds-Averaged Navier-Stokes (RANS) framework to investigate convective heat transfer over rough surfaces with different roughness heights and topology. Several models based on roughness extensions of RANS were compared with the resolving-roughness approach proposed in this study. In comparison with experimental results, the predictions of this approach broadly match the of the velocity profile for a wide range of roughness heights. The proposed approach predicted the expected downward shift in both the velocity and temperature profiles due to roughness, and the variations of these profiles for different roughness topologies.