Worked on enhancing graph visualization and type safety in open-source Python projects. In pyg-team/pytorch_geometric, addressed a longstanding issue where directed edge arrows in graph visualizations did not accurately reflect underlying data, updating the arrowstyle in matplotlib’s FancyArrowPatch to match GraphViz conventions and improve diagnostic reliability. Later, contributed to python/typeshed by refactoring NetworkX type stubs, extracting types from docstrings and updating type annotations to improve code clarity and maintainability. Leveraged skills in Python, Matplotlib, and NetworkX, focusing on targeted, low-impact changes that improved both visual accuracy and type correctness for downstream users and development tooling.