Members of the JUMP Center for Brain-Inspired Computing (C-BRIC) have recently published an exciting new work with SRC member Intel Corporation in the ACM Transaction on Embedded Computing Systems Special Issue on Real-Time Computing in the IOT-to-Edge-to-Cloud Continuum. The work, “PArtNNer: Platform-Agnostic Adaptive Edge-Cloud DNN Partitioning for Minimizing End-to-End Latency," introduces a groundbreaking solution to enhance the efficiency of AI applications, particularly in scenarios requiring real-time responsiveness. 

Lead author Dr. Soumendu Kumar Ghosh graduated with his Ph.D. in 2023 and went on to full-time employment at Intel Corporation. Collaborating on this work were also Vice President for Global Partnerships and Director of Semiconductor Education at Purdue University Prof. Vijay Raghunathan, C-BRIC Associate Director and Purdue University Silicon Valley Professor of Electrical and Computer Engineering Anand Raghunathan,  and 2022 Mahboob Khan Outstanding Liaison award winner Dr. Arnab Raha of Intel Corporation.

PArtNNer addresses the challenge of latency reduction by intelligently distributing AI workloads across devices like smartphones, smartwatches, drones, and cloud servers. Applications such as healthcare diagnostics, industrial automation, and autonomous vehicles could see significant improvement. A standout feature of PArtNNer is that it is platform agnostic. This ensures seamless operation across diverse computing environments, offering adaptability and optimization for various devices. Employing adaptive partitioning, PArtNNer dynamically adjusts workload distribution based on task complexity and available resources, ensuring optimal performance and minimal latency in real-time. Ultimately, PArtNNer signifies a significant advancement in AI optimization, promising more responsive, accessible, and cost-effective solutions in today's interconnected world.

View Dr. Vijay Raghunathan’s JUMP C-BRIC project 2777.018, https://app.pillar.science/projects/3253/overview