The present work shows an innovative design framework for fluid Topology Optimization (TO) able to fully exploit the flexibility offered by Additive Manufacturing (AM) in the production of fluid-structure interaction systems. We present a geometry optimization method able to automatically design complex and efficient heat exchangers, adapted to maximizing fluid-structure heat transfer while minimizing turbulent flow pressure drop. The core of the method is the in-house Fluid Topology Optimization solver extended to include conjugate heat transfer problems. The TO method consists in emulating a sedimentation process inside an empty cavity in which a fluid dynamics problem is numerically solved. A design variable, in this case impermeability, is iteratively updated across the fluid dynamics domain. This mechanism leads to the formation of internal solid structures accordingly to a Lagrangian multi-objective optimization approach, adopted to include a multi-objective function. The solution of the optimization routine is the set of solidified structures, shaping the final optimized geometry. In order to match engineering applications, real conditions are implemented: an impermeability dependent thermal conductivity is included and a smoother operator is adopted to bound numerical thermal conductivity gradients across solid and fluid regions. The optimization is performed on a 3-dimensional straight duct: on the walls the temperature is constant and a coolant turbulent flow is simulated (Re 10000) inside the duct. The solver builds structures enhancing the heat transfer level between the walls of the domain and a coolant flow, by generating counter rotating vortices and complex fluid patterns. This is consistent to solution proposed in the open literature, such as v-shaped ribs, even if the geometry generated is more complex and efficient. The solution is validated with a high fidelity numerical simulation on StarCCM+, using a Detached Eddy Simulation (DES). Validation results shows higher heat transfer efficiency compared to the results present in the literature: the average Nusselt number computed on the domain walls is about 20% higher than the value obtained through experimental investigations on v-shape ribbed ducts. It is the first time that this method is applied and validated on real working conditions.