Developed a multi-strategy pod routing and batch scoring feature for the vllm-project/aibrix repository, focusing on scalable backend infrastructure. The work centered on implementing metrics-based load balancing to optimize resource utilization and throughput across Kubernetes pods. By introducing routing strategies that enable efficient batch scoring, the solution addressed the need for dynamic workload distribution in a microservices environment. The implementation adhered to gateway architecture standards and maintained code quality through disciplined release practices, including signed-off commits. Go was used as the primary language, leveraging expertise in backend development and Kubernetes to deliver a robust, maintainable system for scalable batch processing.