Worked on enhancing distributed multi-node support for vLLM within the ai-dynamo/dynamo repository, focusing on scalable, high-throughput inference across clusters. Integrated Tensor and Pipeline parallelism to enable robust multi-node deployments, addressing both tensor and data parallelism requirements. Improved the distributed execution flow by refining command injection, which reduced setup errors and misconfigurations during deployment. Utilized Go programming and Kubernetes to implement these backend features, ensuring more resilient and stable distributed execution. The work resulted in improved throughput and stability for model inference in distributed environments, demonstrating depth in backend development and distributed systems engineering within a production-grade codebase.