Pranesh contributed to the microsoft/onnxruntime repository by enhancing the CopyTensors API to support asynchronous tensor copying, focusing on improving streaming performance and correctness within the TRT RTX Execution Provider. Using C++ and CUDA, Pranesh addressed a critical issue by ensuring the stream parameter was properly passed, enabling efficient asynchronous operations. Additionally, Pranesh resolved a correctness problem by modifying the createNotification API to pass cudastream by value, eliminating a dangling reference. This work demonstrated a solid understanding of API development and low-level memory management, delivering a targeted feature that improved both reliability and performance for tensor operations in production environments.