Worked on the GoogleCloudPlatform/ml-auto-solutions repository to enhance system stability and deployment reliability. Addressed dependency management by pinning sqlalchemy-spanner to version 1.11.0, preventing compatibility issues with Airflow and related Python packages. Improved the testing infrastructure by updating cluster locations and regions in the map_reproducibility test suite, ensuring tests executed against the correct endpoints. These changes reduced the risk of flaky CI runs and deployment failures, supporting more reproducible results across environments. Focused on DevOps practices and testing, the work emphasized careful dependency control and infrastructure accuracy, contributing to smoother continuous integration cycles and more dependable production deployments.