Worked on the UKGovernmentBEIS/inspect_ai repository to enhance the stability and reliability of data analytics pipelines. Focused on improving the FileRecorder component by implementing a bug fix that ensures empty sample data is handled gracefully, always returning an empty array when no samples are present. This approach reduces downstream errors and supports safer production deployments. Added comprehensive tests for empty sample scenarios and related failure conditions, aligning behavior across Recorder implementations for consistency. Utilized Python for both refactoring and testing, with an emphasis on robust data analysis practices. The work improved predictability and robustness in handling missing or incomplete data.