Worked on backend development for the LianjiaTech/bella-openai4j repository, focusing on enhancing data integrity and compatibility in batch processing with Zhipu AI. Addressed a critical type-safety issue by refactoring timestamp fields in the Batch class from Integer to Long, preventing potential data overflow and ensuring accurate batch processing times. This change aligned the data model with external API responses, improving reliability for downstream analytics. Utilized Java to implement the fix, demonstrating attention to robust data handling and integration requirements. The work centered on bug resolution rather than feature development, reflecting a targeted approach to maintaining backend system stability.