Developed a Kubernetes-free deployment path for the llm-d/llm-d routing stack, replacing InferencePool with a YAML-based endpoint discovery mechanism and a disk-sourced endpoints file monitored by fsnotify. This work delivered comprehensive deployment guidance for EPP, Envoy, and vLLM, targeting NVIDIA GPUs and supporting other accelerators through modelserver overlays. The approach emphasized plain YAML configurations, detailed troubleshooting steps, and architecture diagrams, all reorganized under a consistent router/ documentation structure. Utilizing skills in Cloud Infrastructure, Containerization, and DevOps, the developer incorporated reviewer feedback and improved build-from-source options, resulting in a more accessible and modernized deployment process.