Tritium self-sufficiency is a critical prerequisite for future fusion reactors. The tritium breeding blanket, as the component responsible for in-vessel tritium generation, requires coordinated neutronic and engineering optimization in order to maximize its achievable tritium breeding ratio (TBR). In this work, a high-fidelity 22.5° toroidal sector neutronics model of the China Fusion Engineering Test Reactor (CFETR) equipped with a Helium-Cooled Ceramic Breeder (HCCB) blanket was established. On this basis, we developed a Multi-physics Coupling Intelligent Neutronic Optimization code (MCINO), a two-stage neutronics optimization framework that combines global exploration by simulated annealing with subsequent local refinement. The objective was to maximize the global TBR by optimizing the radial distribution of breeder (Li4SiO4) and neutron multiplier (Be) zones. The optimized design increased the global TBR to approximately 1.193, corresponding to an 8.36% improvement over the initial configuration. The improvement is associated with a more effective radial allocation of breeding and multiplying materials, which enhances neutron moderation, multiplication, and use for tritium production. The optimization workflow was designed to reduce the number of expensive high-fidelity transport recalculations, thereby improving computational efficiency relative to direct brute-force search. Finally, the engineering feasibility of the optimized design was checked through three-dimensional thermal–hydraulic verification, which confirmed that the representative modules remained within their prescribed operating limits. The present work provides an efficient and physically transparent framework for integrated blanket neutronics design and optimization.