Worked on the LearningCircuit/local-deep-research repository to enhance the robustness of citation handling in AI-driven research workflows. Addressed a critical issue where the citation handler would crash if the language model returned a string instead of the expected structured message object during follow-up analyses. Developed a safe processing pathway in Python to handle both string and object responses, improving error handling and reducing runtime failures. This fix ensured uninterrupted research iterations by preventing errors in the citation flow. The work demonstrated strong skills in AI integration and Python development, focusing on reliability and maintainability within complex research automation systems.