Worked on the apache/airflow repository to enhance Spark job management by implementing a feature that automatically deletes Spark jobs failing to start within a configurable timeout. This approach improved reliability and error handling by preventing orphaned resources and increasing failure visibility. The solution was developed using Python, with a focus on backend development and robust error handling practices. Unit tests were added to verify the deletion logic, ensuring correctness and maintainability. Additionally, a minor variable rename improved code readability and clarified intent. The work demonstrated attention to both operational reliability and code quality, with an emphasis on test coverage and maintainable design.