This study introduces an innovative framework that integrates multi-scale planning with multi-objective optimization (MOO) and machine learning to enhance slope safety and geological hazard resilience in hilly smart cities. The framework coordinates city-, district-, and site-level interventions by coupling deterministic slope mechanics with data-driven susceptibility prediction. Empirical results from Nan’an (Chongqing) and Baiyun (Guiyang), the framework integrates deterministic slope mechanics, machine learning–based susceptibility models, and multi-objective optimization (MOO) across city, district, and site levels. Input datasets included 5 m DEMs, 568 documented landslides, 132 triaxial shear tests, and 10,000 stochastic rainfall simulations. Three scenarios, baseline, uniform GI, and targeted GI, were tested under 10, 50, and 100 year rainfall events. Results show that targeted GI allocation raised the factor of safety (FoS) from 1.24 ± 0.19 to 1.67 ± 0.18, reduced Newmark displacement probability from 0.31 to 0.12, and lowered peak runoff from 131.7 to 107.1 m3/s under 50 year storms. Expected annual loss declined from 26.8 to 17.3 million CNY, while ecological connectivity increased from 0.48 to 0.74. Ablation analysis further revealed that infiltration trenches contributed 43.2% of FoS gains, followed by terraced vegetation (31.7%) and check-dams (15.9%). These findings confirm that optimization-driven, cross- scale GI planning significantly outperforms uniform layouts, offering policymakers a robust and transferable pathway for climate-adaptive smart city development.