Developed a rubric-driven GEval scoring normalization feature for the confident-ai/deepeval repository, focusing on backend development and metric implementation using Python. The work introduced dynamic score range determination and strict mode support, enabling accurate normalization aligned with rubric-defined ranges. By refactoring code to handle score_range and score_range_span, the developer improved metric accuracy and facilitated reliable cross-rubric comparisons. The changes addressed previous normalization regressions, resulting in cleaner data and reducing the need for manual adjustments. Throughout the process, clear commit history and maintainable code practices were demonstrated, enhancing the reliability and traceability of evaluation pipelines within the deepeval project.