Worked on reliability and error handling improvements in large-scale data systems, focusing on Apache Avro and apache/incubator-gluten repositories. Addressed a shuffle file permission issue in ColumnarShuffleManager by explicitly setting file permissions to 0644, aligning with Spark’s default behavior and preventing access problems in distributed environments. Enhanced runtime reliability and maintainability by linking changes to tracked issues and providing clear commit documentation. In Apache Avro, improved error messaging for invalid DataFileStream sync markers, clarifying causes such as data corruption or truncation, and updated unit tests to ensure robust error handling. Utilized C++, Java, and test-driven development practices throughout.