Worked on the llm-d/llm-d repository to deliver a new SGLang deployment option for the PD disaggregation service on Kubernetes, enabling SGLang as the inference server. This involved creating and updating YAML configuration files to support scalable, cost-efficient inference for multilingual models. The approach focused on improving resource utilization and inference throughput, aligning with the PD disaggregation well-lit path. Collaboration with reviewers led to refinements in code quality and documentation, including guide corrections and Pod label additions. The work leveraged skills in configuration management, DevOps, and machine learning, laying the foundation for more efficient model deployment workflows in production environments.