Worked on expanding the BerriAI/litellm repository by implementing vision capabilities for Qwen VL models, enabling support for multimodal visual tasks. Focused on AI development and data structure design, the work involved updating the model configuration, specifically the model_prices_and_context_window.json file, to accommodate new deployment parameters and pricing for vision-enabled models. Leveraged JSON for configuration management and modeling to ensure seamless integration of visual features. This addition allows customers to deploy Qwen-VL models with enhanced visual processing, laying the groundwork for future multimodal features and improving the product’s applicability in scenarios requiring both language and vision capabilities.