Worked on the trinodb/trino repository to enhance data reliability by addressing a critical issue in JSON serialization. Focused on preventing non-finite values such as NaN and Infinity from being written to JSON format tables, the developer implemented robust data validation and error handling in Java. This involved introducing a specific error code to surface unserializable values, thereby blocking invalid JSON output and reducing the risk of data corruption or downstream processing failures. The changes improved overall data quality and made it easier for developers to isolate and debug JSON-related issues, strengthening the reliability of JSON serialization workflows.