Worked on the BerriAI/litellm repository to enhance support for large-context enterprise use cases by updating pricing structures and input-token limits for Bedrock-based Claude models. Focused on removing surcharges for context windows exceeding 200,000 tokens and increasing the maximum input tokens to one million for specific models, enabling workflows involving long documents and extended conversations. Ensured pricing consistency across models and regions, providing predictable costs for enterprise clients. Utilized Python and JSON for API development, data management, and testing. Maintained clear documentation and commit traceability to support auditability and future rollbacks, reflecting a methodical and transparent engineering approach throughout the project.