Worked on the EvolvingLMMs-Lab/lmms-eval repository to deliver a forward-compatible OpenAI API integration, addressing upcoming API deprecations and ensuring continued model stability. Focused on backend development using Python, the work involved updating openai_compatible.py to support new API parameters such as max_output_tokens and introducing model-specific adjustments for 'o1' and 'o3'. Added reasoning_effort and response_format controls to align with evolving OpenAI standards, enabling more deterministic and resilient responses. This feature improved the system’s adaptability to API changes, reduced the risk of runtime errors, and preserved downstream functionality, demonstrating strong skills in API integration and backend engineering.