Developed a focused enhancement for the nv-auto-deploy/TensorRT-LLM repository by implementing Python-C++ bindings to streamline LLM argument configuration. Leveraging Python and C++ along with PybindMirror, this work introduced new Python configuration classes for SchedulerConfig and PeftCacheConfig, enabling seamless mapping to their C++ counterparts. The approach improved configuration management by increasing flexibility and maintainability, allowing for faster experimentation and reduced setup risk in LLM workflows. By focusing on API design and configuration management, the solution accelerated feature iteration and delivered tangible value for deployment scenarios, reflecting a deep understanding of cross-language integration and robust software engineering practices.