Artificial intelligence continues to drive global innovation, yet the computation and energy demands of large neural networks strain CMOS hardware due to scaling limits and the memory-processor bottleneck. Neuromorphic computing offers a promising route to energy-efficient, brain-inspired processing. Among emerging device technologies, spintronic elements, particularly magnetic domain-wall and skyrmion devices, magnetic tunnel junctions MTJ, provide non-volatility, high endurance, nanosecond switching, and tunable intrinsic stochasticity desirable for synaptic and neuronal functions. In this talk, I will present my research at the interface of spintronics and neuromorphic computing, aimed at developing energy-efficient hardware intelligence from the device level upward. I will first introduce spintronics as a platform offering non-volatility, intrinsic nonlinearity, and low-energy switching. Then I will discuss spin transport and magnetization dynamics in magnetic tunnel junctions (MTJs), using non-equilibrium Green’s function formalisms coupled with micromagnetics to study MTJ tunnel magnetoresistance and thermal stability, and to examine their scaling with MTJ dimensions. The main part of the talk focuses on my PhD and postdoctoral research work at KAUST, where I progressed from MTJ device physics to spintronic neuromorphic devices and circuits. I will present my research on current and voltage-controlled domain-wall and skyrmion-based synapses and spiking neurons, covering their modeling, fabrication, experimental characterization, and integration into artificial and spiking neural networks. Building on this, I will discuss my current postdoctoral work on field-free stochastic MTJ neurons and multi-state domain-wall MTJ synapses for hardware implementations of Restricted Boltzmann Machines, highlighting how intrinsic thermal noise and spintronic nonlinearities can be used for sampling, unsupervised learning, and optimization. Finally, I will present my research plan to establish a Spintronic Physics for Intelligent Neuromorphic Systems (SPINS) laboratory that integrates materials, devices, circuits, and neuromorphic architectures. The program will focus on the array-level co-design of stochastic MTJ neurons, MTJ spiking neurons, and analog-domain-wall MTJ synapses, integrated on a unified platform. In close collaboration with circuit design groups, this work will enable the development of scalable spintronic AI hardware for edge applications such as smart health and optimization tasks.

Biography:

Dr Aijaz H. Lone is a researcher in spintronics and neuromorphic computing, currently a Postdoctoral Fellow in the Integrated Intelligent Systems (I?S) group at King Abdullah University of Science and Technology (KAUST) under the guidance of Prof. Gianluca Setti. He received his MS in Electrical and Computer Engineering from the Indian Institute of Technology (IIT Mandi) in 2020 and completed his PhD in ECE at KAUST in 2024, focusing on spintronic device physics and hardware-oriented neuromorphic computing. His research spans modeling, simulation, fabrication, and experimental characterization of spintronic devices, including magnetic tunnel junctions (MTJs), domain-wall devices, and skyrmion-based structures. During his PhD, he developed spintronic synapses and spiking neurons and studied their integration into artificial and spiking neural networks at circuit and system levels for energy-efficient computing beyond CMOS. In his postdoctoral work at KAUST, he is advancing field-free stochastic MTJ neurons and multi-state domain-wall MTJ synapses toward hardware implementations of Restricted Boltzmann Machines, targeting ultra-low-power edge AI and optimization applications. Looking ahead, his research vision is centered on establishing the Spintronic Physics for Intelligent Neuromorphic Systems (SPINS) laboratory, aiming to bridge fundamental spintronic device physics with neuromorphic circuits, architectures, and edge AI applications. His work seeks to translate physical spin dynamics into scalable, low-power intelligent hardware.