Yude Levi focused on backend development and pricing configuration for the BerriAI/litellm repository, addressing a critical bug in the pricing registry for GPT-5.x models. By integrating Python-based logic, Yude ensured that the web search capability flag was correctly set for GPT-5.1, GPT-5.2, and GPT-5.3, preventing misalignments between model features and pricing data. This fix improved the reliability of API integrations and reduced the risk of misconfiguration in production environments. Delivered through two traceable commits, the work demonstrated attention to detail and readiness for QA validation, reflecting a targeted and methodical approach to backend stability.