Worked on the yugabyte/yugabyte-db repository to improve the reliability of YSQL major upgrades by addressing issues with stale socket directories that caused retry failures and unpredictable preflight checks. Utilized C++ and backend development skills to implement a ScopeExit guard, ensuring automatic cleanup of temporary socket directories after failed upgrades. Enhanced the upgrade workflow by adding explicit cleanup logs and strengthening test coverage through targeted tests and CI plans. This approach reduced manual intervention and downtime during upgrades, making release cycles more predictable. The work demonstrated expertise in C++ RAII patterns, database management, and continuous integration for robust upgrade processes.