Cloudburst AI: Orchestrating SRE-Driven AI Inference Across Multi-Cloud and Edge Zones

Artificial Intelligence (AI) workloads are increasingly deployed across heterogeneous infrastructures consisting of public clouds, private clouds, and edge computing environments. However, achieving reliable, cost-efficient, and scalable inference across these distributed environments remains challenging. This paper presents Cloudburst AI, a Site Reliability Engineering (SRE)-driven orchestration framework that automates AI inference placement, reliability governance, observability, failure resilience, and operational performance across multi-cloud and edge zones. The framework introduces reliability-centric orchestration policies, autonomous failover intelligence, adaptive routing based on latency and availability, and error-budget driven execution to ensure sustained Service Level Objectives (SLOs). Experimental results demonstrate reduced latency variance, improved availability, minimized failure blast radius, and cost aware inference stability. Cloudburst AI establishes a unified paradigm for AI SRE, bridging DevOps reliability engineering with scalable inference execution.