Diffractive Optical Neural Networks for Intelligent Vision and Computing

About the project

About the Project Artificial intelligence is transforming how machines see and interpret the world, yet most AI systems remain fundamentally electronic: light is first captured, digitised, and only then processed by computational hardware. This project asks a different question: can light itself become the computing medium? This PhD project will explore free-space optical neural networks, where diffraction, interference, and wave propagation are engineered to perform neural-network-like transformations directly in the optical domain. Moving beyond integrated photonics, the project will focus on scalable free-space architectures based on diffractive optics, programmable phase masks, spatial light modulators, metasurfaces, and hybrid optical–digital systems. The central aim is to bridge computational imaging and optical machine learning, developing optical systems that do not merely form images, but actively compute, infer, and extract meaning as light propagates. Potential directions include diffractive neural networks for image classification, optical feature extraction, imaging through complex media, physics-informed optical learning, and hybrid architectures that combine optical front-end intelligence with digital neural networks. The project will combine Fourier optics, computational imaging, machine learning, and experimental optical design. It offers the opportunity to contribute to an emerging paradigm in which vision, sensing, and computation are no longer treated as separate stages, but as a unified physical process enabled by light. Candidates with strong backgrounds in one or more of the following areas are encouraged to apply: • We are looking for a highly motivated candidate with a strong background in physics, optics, photonics, computer science, or a closely related discipline. Prior experience in computational imaging, Fourier optics, wave propagation, machine learning, or optical system design would be advantageous. • The ideal candidate should be curious, mathematically confident, and excited by interdisciplinary research at the boundary between optical physics and artificial intelligence. Experience with numerical simulation, Python, MATLAB, PyTorch, or similar computational tools is desirable. Hands-on experience with free-space optical experiments, spatial light modulators, cameras, lasers, or imaging systems would be a plus, but is not strictly required. • Most importantly, the candidate should be eager to explore a new research direction, think creatively across disciplines, and help build a bridge between computational imaging and optical neural networks.

Research areas