Real-time decision systems increasingly rely on distributed artificial intelligence models deployed across edge devices and centralized cloud platforms. Applications such as autonomous monitoring, industrial automation, smart cities, and intelligent transportation require ultra-low latency inference while maintaining scalable model management and coordination. This paper proposes a hierarchical edge-to-cloud AI orchestration framework that dynamically allocates inference tasks, manages model synchronization, and optimizes latency–cost tradeoffs. The architecture integrates edge inference agents, regional aggregation nodes, and cloud-based control intelligence to enable scalable and reliable real-time decision support. Experimental evaluation demonstrates latency reduction, bandwidth efficiency, and improved decision consistency compared to cloud-only architectures. The proposed framework establishes a scalable foundation for distributed AI orchestration in latency-sensitive environments