Abdoulaye Ndao's lab at UC San Diego has cracked a problem that has plagued microscopes, telescopes, and smartphone cameras for decades: the subtle blur caused by lens imperfections that distort light and degrade images. Using a combination of artificial intelligence and a meticulously engineered optical device no larger than a postage stamp, the team has developed a way to detect and correct these distortions from a single image—a breakthrough that could transform how advanced optical systems are built and used.
Blurry light from lens aberrations has long required multiple measurements, extra hardware, and complex calculations to address. This made optical correction systems bulky, slow, and difficult to integrate into compact devices. The UC San Diego approach sidesteps all of that. The team paired an artificially designed optical element—a metasurface made of tiny titanium dioxide nanopillars arranged on glass—with a deep neural network trained to read distortions instantly. The result: the system can now identify aberrations from a single snapshot of the light pattern and automatically determine how to correct them.
"We used a combination of fundamental physics, nanofabrication and machine learning to make hidden distortions easier to detect and correct," said Ndao, an electrical and computer engineering faculty member in the Jacobs School of Engineering and affiliate of the Qualcomm Institute. "Our fast, robust solution is tiny and easy to integrate into different optical systems. The weight is almost nothing, because the size of the sample can be one by one centimeter and half a millimeter thick."
The work, co-led by Ph.D. students Sina Moayed Baharlou (also of Boston University) and Muhammad Waleed Khalid, wasn't theoretical. The researchers fabricated the optical components at UC San Diego's Nano3 cleanroom facility and tested the approach experimentally across multiple wavelengths, noisy conditions, and complicated beam shapes. They kept expanding the device's capabilities until it handled real-world complexity. The work was published in Nature Communications in May 2026.
What makes this approach genuinely scalable is its simplicity: no repeated calculations, no bulky apparatus, no need for multiple measurements. A single image is enough. The AI does the heavy lifting, trained to recognize the unique signature each type of distortion leaves on the light pattern. This shift from hardware-heavy correction to intelligent, single-shot detection could reshape how optical systems are designed and deployed.
The implications ripple across fields that depend on pristine imaging: biology and microscopy, where sharper images mean clearer views of cellular structures; astronomy, where correcting telescope distortions improves observations of distant objects; and precision manufacturing, where optical accuracy is critical. For each field, the promise is the same—faster systems, smaller devices, easier integration.
The patent-pending approach represents what Ndao calls a "scalable and practical foundation for real-time aberration correction for next-generation optical and photonic systems." A postage-stamp-sized device that weighs almost nothing, paired with machine learning, has turned an old problem into a solved one. The result is an optical system that sees more clearly—and does so instantly.