Stanford University AI platform accelerates optics design

23 Dec 2025

New computational tools and AI assistants set to benefit optical computing and metasurface devices.

Stanford University has introduced a new framework of computational tools and AI assistants able to speed up the design of freeform metasurfaces and associated optical systems.

The new toolbox has been named MetaChat, and is intended to allow real-time collaborations using self-reflective AI agents and fast simulations.

Described in Science Advances, Stanford's work will streamline complex design tasks and reduce the burden on optical designers, in order to support the creation of advanced photonic systems.

MetaChat is a multi-agentic framework, applying an iterative process to interface AI agents with code-based tools, other specialized agents and human designers. It aims to translate semantically described photonic design goals into high-performance, freeform device layouts.

"This combination of agentic AI with high-speed surrogate computational models is new," said Stanford's Jonathan Fan. "We have multiple agents talking with those tools and the user to do complex design tasks. That's a really big opportunity."

In designing MetaChat, Stanford first built a deep-learning neural network able to solve Maxwell's equations governing electric and magnetic fields, and then created AI agents to play the roles of optics designers and materials experts. To improve decision-making, the team used prompts to give the AIs agency, including the ability to self-reflect. MetaChat combines these computing tools and AI agents together, with a chat interface for users to make design requests.

Metasurfaces are one technology where MetaChat could bring significant benefits. The fine-scale designs needed to exploit the optical properties of metasurfaces require simulations that model how electrical and magnetic fields are produced and change over time, and the simulations must be run thousands of times as the design undergoes trial and error.

Rapid AI collaborations preserving human insights

"It really snowballs into something that takes on the order of weeks or even months to do for large devices," commented Stanford's Robert Lupoiu.

MetaChat could simplify the process, as when the Stanford project tasked it with designing a lens that can simultaneously focus blue light to one point, and red light to another point. The materials expert AI agent queried a database to identify materials with the properties needed, before the AI designer configured the tiny blocks and pinged Lupoiu with clarifying questions.

As a result the project found it took MetaChat 11 minutes to produce a downloadable design that was comparable to state-of-the-art devices.

The Stanford team predict that similar systems with autonomous AI agents could accelerate technology in other areas of optical technology, where researchers could develop their own specialized, self-reflective AI agents. This could potentially allow rapid cross-disciplinary collaboration.

"There's a real shortage of optical designers," said Jonathan Fan. "There's a huge need for people to build various types of photonic systems, so I think this is something that can really help with that. But such platforms won't make humans obsolete. The goal is to utilize insights from people. It takes a person to ask the right questions and to identify when something is not right."