Abstract: Hydrodynamic retarders are often required for the three-dimensional flow behavior. Specifically, the conventional straight blade configurations are limited to low braking performance and geometric adaptability. In this study, a three-dimensional geometric modeling was proposed for the cambered blade retarders using the class-shape-transformation (CST) approach. The toroidal circulation path, blade camber line, and thickness distribution were all parameterized using CST curves, particularly for the highly flexible and precise representation of the blade geometry. A comparative analysis was conducted to verify the reliability and accuracy of the computational approach. A comparison was also made on the experimental braking performance data and numerical simulations using computational fluid dynamics (CFD) with the baseline model of a straight blade retarder. The results revealed that there was a maximum deviation of 4.92% and a minimum deviation of 4.43% between the simulated and experimental values, both of which fell below the generally accepted threshold of 8%. The CFD model was also verified for the subsequent optimization. An improved cambered blade retarder was reconstructed to combine the simulation framework and CST-based parameterization. A CFD analysis was then conducted between the straight and cambered blade configurations. The cambered blade retarder fully met the feasibility requirements from a structural standpoint. The braking performance also outperformed the conventional straight blade variant. Both the structural viability and performance of the cambered blade were implemented in the retarder applications. The braking torque was reduced due to the idling loss of power. A design of experiments (DOE) was employed to systematically explore the influence of the key blade cascade parameters on the retarder performance. A response surface model (RSM) was developed to quantitatively describe the relationship between the cambered blade cascade parameters and two critical performance metrics: braking torque and idling loss power. A main effects analysis was then conducted on the RSM in order to identify the individual contributions of each design parameter. The analysis revealed that the peak height of the blade camber line had a significant positive impact on the retarder performance, thus enhancing the braking torque to mitigate the idling loss of power. Conversely, the blade deflection angle, incidence, and thickness factor were all found to exhibit negative correlations with the overall retarder efficiency. These variables were also optimized to avoid performance degradation. The optimal combination of the blade parameters was determined to balance the performance. A multi-objective evolutionary optimization was utilized as the non-dominated sorting genetic algorithm II (NSGA-II). The NSGA-II optimization generated a set of optimal solutions, from which an optimal blade configuration was selected using design priorities. An optimally cambered blade retarder was evaluated to compare with the pre-optimization, in terms of its external performance and internal flow field. The braking performance of the optimal blade retarder was 23.5 % higher than that of the original. The idling loss of power was reduced by 30.9 %. In conclusion, a robust CST-based approach was presented for the geometric modeling and performance optimization of 3D cambered blade hydrodynamic retarders. Advanced parameterization techniques were integrated to validate the CFD analysis and evolutionary multi-objective algorithm. The finding can offer a comprehensive framework for the high-performance retarders. Beyond the immediate application to hydrodynamic braking systems, the modeling and optimization techniques can provide a valuable reference in the broader field of turbomachinery. Some insights and tools can also be extended to improve the efficiency of the turbines and rotating fluid machinery.