Worked on stabilizing the test infrastructure and database layer for the NevaMind-AI/memU repository, focusing on reliability across different environments. Addressed two critical bugs by improving the robustness of SQLite test suite file path handling and updating the PostgreSQL MemoryItemModel schema to allow null resource identifiers. These changes reduced test flakiness in continuous integration and prevented database errors when handling memory items without associated resources. Leveraged skills in Python, PostgreSQL, and database modeling to deliver well-documented, targeted fixes that enhance code quality and maintainability. The work emphasized cross-backend consistency and laid a foundation for more reliable future development and testing.