Worked on the intel/onnxruntime repository to deliver 64-bit UDMA mode support for the QNN execution provider targeting architecture version 81 and above. This feature involved implementing new configuration options and updating backend methods in C++ to enable efficient model execution on compatible Intel hardware. The technical approach focused on performance optimization, allowing higher data throughput and improved efficiency for QNN-based workloads. No bugs were addressed during this period, with efforts concentrated on feature development. The work demonstrated a strong understanding of backend development and hardware-specific enhancements, resulting in direct performance benefits for users running models on supported devices.