Abstract: The thermo-fluid behavior of reactor core coolant flow channels plays a decisive role in determining the overall heat removal capability and safety of nuclear systems. In the context of next-generation nuclear reactors, optimizing the structural configuration of flow channels has emerged as a critical challenge, with the goal of enhancing thermal performance, improving operational reliability, and reducing engineering costs. This study aims to explore a novel approach for the intelligent design of reactor flow channels by integrating advanced topology optimization with parametric generation algorithms under coupled physical field constraints. In this study, a multi-physics optimization framework was developed and implemented. The topology optimization process was carried out using the COMSOL multiphysics finite element platform, where the fluid-solid thermal interaction was modeled through conjugate heat transfer simulations. The transformational challenge reactor (TCR) was selected as a representative case for applying the proposed methodology. A surrogate-free, heuristic evolutionary algorithm based on the Nelder-Mead simplex method was adopted to iteratively optimize channel morphology without relying on gradient information. Both traditional parametric design and intelligent generative models were constructed and compared to evaluate the benefits of automated geometry evolution. The optimization problem was formulated to minimize the peak temperature within the core while preserving acceptable pressure drops, thereby ensuring practical engineering applicability. The optimized solutions consistently converge toward a fractal “Y”-shaped bifurcation pattern, characterized by enhanced surface area distribution and minimized thermal resistance. This configuration draws inspiration from natural branching systems and exhibits superior thermal performance compared to conventional circular flow channels. Quantitative results show that the intelligently optimized channel geometry reduces the maximum core temperature by 127.6 K while keeping the pressure drop within acceptable design limits (＜10 kPa). This work confirms that intelligent topology optimization can serve as a powerful tool for reactor cooling system design. It contributes a scalable and generalizable methodology for handling coupled multi-physics problems in complex domains. By merging evolutionary algorithms with high-fidelity simulations, the proposed approach lays a robust foundation for future developments in intelligent reactor design, with implications extending beyond nuclear power systems to broader heat transfer applications in aerospace and energy industries.