Max Robotics — Global Service Robot Directory + US Certification

About

Seulki Lee is a researcher working at the intersection of deep learning systems, embedded computing, and intelligent robotics. His work primarily focuses on making deep neural networks (DNNs) practical and efficient for resource-constrained environments, with a particular emphasis on real-time inference, memory optimization, and model interpretability. Lee's most influential contributions include SubFlow (2020, 47 citations), a dynamic adaptation strategy that enables real-time DNN inference and training under fluctuating timing constraints, and Neural Weight Virtualization (2020, 41 citations), a novel approach that allows multiple DNN models to be packed into fixed-size memory on embedded systems, enabling fast and scalable multitask learning without the typical memory bottlenecks. Together, these works address critical deployment challenges that have long hindered AI adoption on edge devices. His Deep Functional Network (DFN) framework further pushes the field forward by offering semantic interpretability of DNNs through functional program approximations, bridging the gap between black-box models and explainable AI. More recently, Lee has expanded his research scope to include human-robot collaboration in complex indoor environments, co-authoring a 2025 review examining multi-robot systems in construction settings. His diverse and growing body of work reflects a commitment to translating advanced AI research into real-world, safety-critical applications.