• Proposes an end-to-end framework for designing, simulating, and iteratively optimizing a UAM corridor along Brasilia's Eixão, aligned with Brazil's BR-UTM ecosystem.
• LLM-assisted optimization with GPT-4o Mini and Gemini 2.5 Pro enables root-cause diagnosis and rapid prototyping of GA variants, achieving a 59.62% reduction in maximum waiting time.
• GA v5 approaches the robustness of round-robin scheduling, while poorly designed fitness functions in GA v2–v4 and GA v6 degrade performance.
• Demonstrates that generative AI can accelerate simulation and co-create operational logic, offering a replicable blueprint for early-stage UAM planning.