Worked on the onnx/onnx repository to enhance the robustness and correctness of the ONNX Extractor’s subgraph extraction workflow. Addressed a key issue by ensuring that empty input and output names are skipped during graph processing, which prevents the creation of incorrect subgraphs when models contain optional or unused inputs and outputs. This fix, implemented in Python and leveraging skills in graph processing and model extraction, improved data integrity and reduced downstream errors in model export and conversion pipelines. The changes contributed to smoother integration with external systems and enhanced the reliability of ONNX model extraction workflows for future development.