Worked on enhancing PPTX document conversion robustness in the DS4SD/docling repository, focusing on backend development with Python. Addressed issues with non-standard and malformed shapes by introducing safe handling for unrecognized shape types, allowing text and table extraction to continue even when encountering problematic elements. Extended error handling to gracefully bypass malformed picture shapes, ensuring the conversion process does not fail entirely. Added regression tests and warning-based diagnostics to improve reliability and surface issues for future improvements. Maintained code quality through lint cleanups and stability enhancements, demonstrating a methodical approach to error handling and unit testing in document processing workflows.