Contributed to data preprocessing and prompt engineering across two open-source projects. Developed the DocumentLineDeduplicator operator for modelscope/data-juicer, enabling scalable, memory-efficient removal of boilerplate lines from large document collections using Python and configurable skip rules to preserve essential content. Enhanced data quality and streamlined downstream analytics by implementing a two-phase deduplication pipeline with parallel processing. In run-llama/llama_index, improved advanced prompt documentation by correcting example errors, updating kernel information, and refining prompt formatting to ensure reliable execution. Demonstrated strengths in Python programming, documentation, and unit testing, with a focus on maintainability, onboarding, and reducing support overhead for developers.