Hafezi Part of Winning Team in UMD Invention of the Year Competition
May 14, 2026
Mohammad Hafezi, a senior investigator in the NSF Quantum Leap Challenge Institute for Robust Quantum Simulation (RQS), is part of an interdisciplinary research team recognized for developing a new optical computing technology designed to improve the speed and efficiency of artificial intelligence systems.
The team won the quantum category of the University of Maryland’s Innovate Maryland Invention of the Year competition for its Topological Photonics Architecture for Optical Computing and Artificial Intelligence, or TOPAI, platform.
In addition to Hafezi, the team includes Mahmoud Mehrabad, formerly at UMD and now a research scientist at the Massachusetts Institute of Technology, Lida Xu of the UMD Department of Physics, and Joint Quantum Institute researchers Supratik Sarkar and Zhi-Yuan Wei.
TOPAI was developed to address growing challenges facing conventional AI computing systems, which consume large amounts of electricity, generate significant heat and become increasingly difficult to scale as demand for computing power rises. Instead of relying on electronic signals, the technology processes information using light.
The platform is built around a topological photonics architecture that uses specially engineered photonic structures to guide light in stable and resilient ways. This field was pioneered by Hafezi more than 15 years ago at UMD. Now, researchers say the system can carry out AI operations such as matrix multiplication, convolution and neuromorphic inference directly in the photonic domain while reducing sensitivity to noise, fabrication imperfections and thermal fluctuations.
Like a train constrained to its tracks, TOPAI’s topological photonic states are designed to keep information moving reliably even as computing networks become more complex. The researchers say the approach could support faster, more energy-efficient AI infrastructure and high-performance computing systems.
The invention combines engineered photonic lattices and waveguide networks that support defect-immune light transport and programmable optical pathways. According to the team, the architecture is compatible with existing silicon photonics and CMOS manufacturing platforms, potentially simplifying large-scale adoption.
Conventional photonic computing systems have historically faced challenges involving instability and scalability, while electronic AI accelerators continue to encounter power and bandwidth limitations.
The researchers say TOPAI helps bridge that gap by creating scalable, modular photonic circuits capable of supporting AI and optical computing workflows without extensive electronic correction systems.
Innovate Maryland annually highlights inventions and startup activity emerging from UMD laboratories, with a focus on technologies that demonstrate commercial and societal potential.
