Worked on the modelscope/data-juicer repository to optimize memory usage during data processing, specifically targeting the convert_to_absolute_paths function. The approach involved refactoring path handling logic to use generator-based processing, which reduced peak memory consumption and improved throughput when handling large datasets. By focusing on Python performance optimization and memory profiling, the changes enabled more scalable and efficient data pipelines, lowering resource costs and supporting faster processing. The work demonstrated a strong command of Python, data processing techniques, and memory optimization strategies, resulting in a feature that enhances the repository’s ability to process data samples efficiently without introducing new bugs.