Turbulent flow and heat transfer over rough and structured surfaces are important in industrial cooling systems, heat exchangers, gas-turbine cooling passages, hydrogen and sustainable-fuel technologies, and additively manufactured thermal components. Surface roughness can enhance heat transfer but generally also increases frictional drag and pumping-power requirements. Reliable prediction of this balance is difficult because momentum and thermal transport depend not only on an average roughness height but also on surface spacing, slope, anisotropy, statistical distribution and spatial organization. This project will use GPU-accelerated lattice Boltzmann simulations to investigate turbulent momentum and heat transfer over selected canonical and engineering-relevant rough surfaces. The requested allocation represents the first stage of a longer computational programme. It will establish and validate the GPU workflow, quantify performance and scalability on NAISS resources, and generate a targeted high-fidelity dataset covering the most scientifically important combinations of roughness geometry, Reynolds number and thermal-transport parameters. The simulations will be performed using an in-house lattice Boltzmann code developed by Kuwata and Suga and further adapted for the present research. The lattice Boltzmann method is attractive for complex rough surfaces because it operates on a regular Cartesian lattice and uses predominantly local numerical operations. The project will first validate smooth-wall and canonical rough-wall turbulent-channel cases against established reference results. A selected set of rough-surface simulations will then be performed to examine the effects of roughness height, spacing and morphology on frictional drag, turbulence statistics and wall heat transfer. The simulations will generate three-dimensional velocity and temperature fields, wall shear stress, pressure-loss data, local wall heat fluxes and statistically averaged turbulence quantities. The results will be used to determine which surface characteristics control the balance between heat-transfer enhancement and hydrodynamic penalty. The project will also provide measured GPU performance, memory requirements and multi-GPU scaling data needed to design a subsequent expanded computational campaign. The work supports the COOLPRO, COOLHEAT, HYDROFUEL and AdTherM research programmes at Lund University and contributes to the development of more energy-efficient thermal-management and energy-transformation technologies.