Worked on enhancing GenAI instrumentation accuracy within the Arize-ai/openinference repository, focusing on backend development and data processing using Python. Addressed a critical issue by correcting the mapping of token usage metrics for cached tokens in Google GenAI instrumentation, ensuring that cached content tokens are now included in total token counts. Developed and integrated automated tests to validate this new behavior, which helps prevent metric drift and supports ongoing reliability. This update improved the precision of usage metrics, enabling more accurate cost accounting and fostering greater trust in GenAI telemetry for customers. Emphasized thorough testing and robust backend practices.