KAIRO: A Browser-First Cloud Operating System for Resource-Constrained Computing Devices

Resource-constrained computing devices limit users who must execute computationally intensive workloads such as machine-learning model training, large-scale software compilation, and media processing. This paper presents KAIRO, a browser-first cloud operating system that relocates such workloads from the client device to remotely provisioned, container-isolated cloud infrastructure while the browser acts as the sole client runtime. We describe the layered system architecture, a working prototype built on Next.js, Express.js, and Docker, and an experimental evaluation across six representative workloads and six network conditions. Relative to local execution on the same device, KAIRO reduced client CPU utilization by 85–95% and client memory utilization by 75–90%, at the cost of a 3–12% increase in end-to-end workload execution time attributable to network communication overhead. Workspace provisioning completed in 8–45s depending on the requested resource profile, and response times remained below 150ms up to 50 concurrent simulated users, degrading to 200–400ms at 200 users on a single node.