Developed the initial TensorRT-based Execution Provider for RTX GPUs within the ROCm/onnxruntime repository, focusing on accelerating machine learning inference and enhancing usability on NVIDIA hardware. The work involved C++ development and CUDA integration to establish a GPU-accelerated execution path in ONNX Runtime, laying the groundwork for future optimizations and broader hardware support. Efforts centered on feature implementation and seamless integration rather than bug fixes, demonstrating disciplined code practices and collaborative development. By leveraging TensorRT and machine learning expertise, this contribution advanced the product roadmap, improved performance, and expanded deployment possibilities for ONNX Runtime on modern GPU architectures.