Worked on the RosettaCommons/foundry repository to enhance analytics metrics logging by introducing robust NaN handling and data cleaning techniques. Refactored the metrics callback in Python to skip logging metrics composed entirely of NaN values and implemented logic to drop columns with all-NaN entries, resulting in clearer and more reliable metrics reporting. Addressed a typo in the validation progress calculation, improving the accuracy of progress tracking. These changes reduced log noise and improved observability, enabling faster issue diagnosis and more trustworthy dashboards. The work demonstrated a focus on data quality, logging best practices, and careful attention to detail in implementation.