Worked on the traP-jp/traO-Judge-judge repository to modularize the output parsing flow and improve the maintainability of the judging pipeline. Developed a dedicated output_parser module in Rust, moving parsing logic out of judge_output.rs to enhance code organization and support future scalability. Focused on backend development and code formatting, the changes aimed to reduce the risk of regression when updating parsing logic and to accelerate future feature work. The refactored structure in judge_core improved testability and made onboarding easier, laying a foundation for safer refactors and more reliable feature development within the Rust-based backend system.