Soroush Heidari

About

Soroush Heidari is a researcher at the forefront of efficient machine learning deployment on edge computing systems. His work centers on the critical challenge of optimizing deep neural network (DNN) execution on resource-constrained edge multiprocessor system-on-chips (MPSoCs), where heterogeneity in both models and hardware demands intelligent resource management. Heidari’s major contribution, exemplified by his highly cited paper “CAMDNN: Content-Aware Mapping of a Network of Deep Neural Networks on Edge MPSoCs” (2022, 5 citations), introduces a novel framework that leverages content awareness to dynamically map networks of DNNs onto available processing elements. This approach maximizes hardware utilization while maintaining inference accuracy, addressing a key bottleneck in real-world edge AI applications. By tackling the complexity of co-scheduling multiple, interdependent neural networks, Heidari’s work enables more efficient and scalable deployment of ML workloads in latency-sensitive environments like autonomous systems and smart devices. His research has already garnered attention for its practical impact on bridging the gap between advanced AI models and the limited resources of edge platforms, establishing him as a promising voice in the intersection of embedded systems and machine learning.

Research Focus

Key Achievements

Top Papers

1. 1
   CAMDNN: Content-Aware Mapping of a Network of Deep Neural Networks on Edge MPSoCs

   5 citations · 2022

Key Collaborators

Contact & Links