Worked on the BerriAI/litellm repository to enhance PDF handling and translation pipelines, focusing on robust support for document processing across Bedrock and Anthropic APIs. Leveraged Python for backend development, implementing stricter MIME type validation and improved error handling to reduce API rejections related to file inputs. Expanded automated test coverage by mirroring PDF tool-result scenarios into unit tests, ensuring parity between integration and CI testing. Incorporated review-driven refinements to edge-case handling, particularly for file data and identifier management. This work strengthened regression protection, improved reliability of document translation results, and demonstrated proficiency in API integration, testing, and backend engineering.