Contributed to the IBM/unitxt repository by enhancing the Schema Linking Task evaluation pipeline, focusing on the addition of precision and recall metrics to provide more detailed insights into model performance. Leveraged Python and data analysis skills to implement these metrics in a maintainable and testable manner, ensuring that the evaluation process now supports more granular benchmarking and targeted model tuning. The work was delivered through clean, well-documented commits, with all changes isolated to the evaluation components. No bugs were reported or fixed during this period, and the technical approach emphasized clarity, reproducibility, and actionable feedback for machine learning practitioners.