Worked on the apple/axlearn repository to address a critical issue in the language-modeling input processor, where an infinite dataset loop could cause training jobs to hang indefinitely. Applied Python and data processing expertise to identify and resolve the bug, ensuring robust handling of infinite datasets during model training. Developed and integrated a regression test using unit testing practices to validate the fix and prevent future regressions. This work improved the reliability of long-running machine learning workflows by reducing wasted compute and strengthening continuous integration coverage for dataset handling, contributing to more stable and maintainable data pipelines within the project.