Worked on improving data correctness in the apache/datafusion-comet repository by addressing a critical issue in Spark expression evaluation. Focused on the spark-expr module, the developer fixed a bug in the IfExpr logic to ensure nullability is accurately determined for both true and false branches during conditional data processing. This change enhanced the reliability of Spark’s conditional logic, reducing the risk of downstream null-related errors in data pipelines. The work involved deep understanding of data processing and expression evaluation, utilizing Rust to implement and validate the fix. The contribution prioritized correctness and robustness in complex data transformation workflows.