Osteochondral defects require biomimetic scaffolds. These scaffolds must simultaneously reproduce the mechanical heterogeneity of subchondral bone and maintain stable integration between bone and cartilage. In this study, we developed a data-driven design and validation framework for patient-specific biphasic Ti6Al4V osteochondral scaffolds using public computed tomography (CT) datasets and finite element optimization. CT images from the Osteoarthritis Initiative database were analyzed to extract Hounsfield unit (HU) distributions and derive anisotropic mechanical parameters through density–elasticity mapping, yielding three representative design targets corresponding to low-, medium-, and high-density cancellous bone. Gradient Gyroid scaffolds with region-specific porosity and strut parameters were subsequently designed to match these physiological targets. To improve osteochondral integration, an interlocking interface architecture incorporating peg-based mechanical anchoring was optimized using parametric finite element response surface modeling and Bayesian optimization. The scaffolds were fabricated via selective laser melting using Ti6Al4V powder and comprehensively characterized through quasi-static compression, hemispherical indentation, and interface shear testing. The fabricated scaffolds exhibited high geometric fidelity, with average dimensional deviations of less than 0.15 mm. The measured macroscopic compressive moduli demonstrated excellent manufacturing fidelity, deviating by less than 10% from the CT-derived design targets across all density groups, thereby offering a viable path to mitigate clinical stress shielding. Furthermore, the optimized interlocking interface significantly increased both indentation stiffness and destructive interface shear strength relative to smooth-interface controls (p < 0.01), closely matching finite element predictions of minimized stress concentration at the osteochondral junction. Taken together, these findings demonstrate that CT-informed, topology-optimized Ti6Al4V scaffold architectures can achieve precisely matched porous mechanics while substantially expanding interface stability, providing a highly reproducible and translatable strategy for personalized load-bearing joint repair applications.