Corrosion-induced material loss in steel bridge poses persistent persistent challenges for inspection, load rating, and rehabilitation, often leading to conservative decisions and labor-intensive repair strategies. This paper presents a cyber-physical workflow for optimized repair design and a robotic cold-spray deposition architecture targeting corroded steel bridge girders. The framework integrates laser scanning, corrosion mapping, nonlinear finite element analysis, and gradient-based optimization to generate material-efficient cold-spray repair geometries tailored to the as-is condition of deteriorated members. Three-dimensional point cloud data are processed into a structured thickness field that captures localized corrosion while remaining computationally efficient for iterative optimization. Using this representation, spatially varying cold-spray deposition thickness fields are determined to maximize load-carrying capacity recovery while minimizing added material. Both Pareto-based and penalty-based optimization formulations are explored, enabling efficiency-driven trade-off analysis or direct targeting of prescribed capacity levels. The computational framework is validated against full-scale experimental testing of a naturally corroded steel girder, demonstrating close agreement between predicted and measured structural response. To connect optimized repair design with execution, a robotic cold-spray deposition architecture and a dedicated slicing strategy are introduced, together with a virtual environment for simulating deposition kinematics and process constraints. The proposed workflow establishes an integrated, data-driven pathway toward automated, performance-informed cold-spray repair of steel bridge infrastructure.