Worked on improving instrumentation reliability for the Arize-ai/openinference project, focusing on backend development and data instrumentation using Python. Addressed a specific issue in the OpenAI API integration by correcting the span kind returned for GuardrailSpanData, which previously led to inaccurate data representation in monitoring dashboards. The solution involved a targeted bug fix with minimal code changes, ensuring that the instrumentation now provides reliable and precise data for observability tools. This work enhanced the quality of monitoring outputs without introducing new features, demonstrating a disciplined approach to debugging and patching within a complex codebase while maintaining alignment with project reliability goals.