Developed a GPU-accelerated ground segmentation module for point cloud processing in the vish0012/autoware.universe repository, focusing on real-time perception for autonomous systems. Leveraged CUDA and C++ to implement high-performance processing nodes, refactored the existing pipeline for improved throughput, and introduced scalable architecture to support efficient resource allocation. Enhanced maintainability by restructuring core modules, improving parameter handling, and integrating pre-commit and CI workflows. Added debugging features such as ground pointcloud publishing and processing-time metrics to facilitate diagnostics and performance tuning. Addressed operational stability by resolving launch issues and input alignment, resulting in a robust, maintainable, and efficient perception component.