Worked on the maximhq/bifrost repository to deliver two core features focused on AI provider integration and cost management. Developed a production-ready Hugging Face Inference Provider integration, enabling chat and text-to-speech capabilities while enhancing test coverage and performance visibility. Introduced a universal CountTokens API across multiple providers, allowing users to estimate token costs and validate input lengths without generating responses. Improved testing reliability by migrating vision-model tests to Cohere Aya Vision and updating CI pipelines and documentation. Leveraged Go for backend and API development, emphasizing robust testing and infrastructure updates to ensure stable, maintainable, and cost-aware AI service integrations.