During August 2025, contributed to the apple/axlearn repository by developing a dedicated Kubernetes service for the Leader Worker Set (LWS), focusing on scalable exposure and streamlined management. Leveraging Python and Kubernetes, implemented a configurable service that centralizes LWS network exposure, allowing for flexible adjustment of service types, ports, and protocols without requiring code changes. This approach reduced operational overhead and simplified future modifications to LWS deployment across environments. The work emphasized API development and cloud computing best practices, with no major defects reported, reflecting a focus on infrastructure enhancement and maintainable deployment patterns within a production-grade Kubernetes environment.