Max Robotics — Global Service Robot Directory + US Certification

About

Shahriar Nirjon is a computer scientist whose research sits at the intersection of embedded systems, real-time computing, and deep learning, with a particular focus on making artificial intelligence practical and efficient on resource-constrained devices. His most impactful contributions center on optimizing deep neural network (DNN) execution for real-world deployment. His SubFlow framework (2020, 47 citations) introduced a dynamic strategy enabling real-time DNN inference and training under shifting timing constraints — a significant advance for latency-sensitive applications. Complementing this, his Neural Weight Virtualization work (2020, 41 citations) tackled the challenge of running multiple deep learning models simultaneously on memory-limited embedded hardware, pioneering scalable in-memory multitask learning. Beyond neural network efficiency, Nirjon has explored novel sensing modalities, including SuperRF, which leverages low-cost mmWave radar to construct enhanced 3D scene representations without cameras — valuable for privacy-sensitive environments — as well as WiFi-based non-line-of-sight human detection for safer human-robot interaction. His more recent Deep Functional Network concept aims to bring interpretability and deployment flexibility to otherwise opaque DNNs. Collectively, Nirjon's work reflects a sustained commitment to bridging the gap between powerful AI models and the practical constraints of real-world embedded systems.