Global Launch of MatterSeek: Redefining the Materials R&D Paradigm with a Focus on Industrial Viability

Yuyao Technology has opened its AI materials design platform to researchers worldwide, together with a closed loop that links performance targets to synthesis routes, simulation checks and laboratory validation.

LOS ANGELES, Nov. 16, 2026 /PRNewswire/ — Yuyao Technology today announced the global launch of MatterSeek, a general-purpose AI platform for the research, design and development of new materials.

Researchers worldwide can now use the platform through its international website and desktop client, with a simultaneous rollout on Hugging Face, LinkedIn, Facebook and X. A beta version has also opened in China, where users can try the platform online at no cost and explore the full range of MatterSeek’s forward- and reverse-design material models.

Materials R&D remains one of the hardest problems in AI for Science. The difficulty is not predicting a single property, but the deep coupling between composition, microstructure, processing and end-use performance. R&D teams work inside a high-dimensional design space, where they must cope with the “curse of dimensionality” and reconcile metrics that pull against one another to find a Pareto-optimal answer.

Two technical routes dominate the field today. The first is simulation-driven, built on density functional theory (DFT) and interatomic potentials. Starting at the atomic level, it explains microscopic mechanisms convincingly, but error accumulates as the analysis moves from the micro to the meso and macro scales, so it struggles to deliver a process plan anyone can implement. The second is data-driven, with models trained on published literature and public databases. It is strong at high-throughput screening, yet its output generally stops at the level of chemical formula and crystal structure. It takes no account of how a material actually behaves in service, and the candidates it surfaces are frequently difficult to synthesize or verify in the lab.

Yuyao Technology’s response to both gaps is a closed-loop R&D paradigm: work backward from a macroscopic performance target to generate material designs, validate them with multiphysics simulation, close the remaining data gaps through targeted experiments, and resolve complex Pareto trade-offs with multi-objective optimization.

MatterSeek is a non-language physics foundation model developed entirely in-house. It supports both the forward and the inverse design workflow. In forward design, a user enters a material composition and receives high-confidence predictions across the full composition–structure–process–performance chain. In inverse design, a user enters target performance metrics and the platform proposes a new material system: its composition, crystal structure, processing route and expected performance.

The inverse design engine can optimize 190 material metrics, drawing on an industrial-grade performance framework of nearly 200 metrics that covers service performance, manufacturability, interfacial properties and weatherability, among others. Any metric can be set as a primary or secondary objective, or used as a constraint, which lets the engine search for the best solution across multi-parameter ranges. Critically, the platform natively returns an executable synthesis route together with the equipment it requires, the operating steps and an estimate of raw-material cost. It goes beyond judging whether a crystal structure is theoretically feasible and answers the engineering questions that follow: how the material is made, and at what cost.

For high-value candidates, the platform can call on DFT, phase-field and finite-element simulation tools as needed to verify results in detail, raising confidence in a prediction while keeping compute consumption under control. Its training data rests on high-quality structured datasets and several hundred thousand real experimental records. That data was not passively collected: experiments were run specifically to close identified R&D gaps. Performance targets, simulation checks and process plans therefore connect into a single R&D chain built for industrial use.

The physical experiment loop is the heart of that chain. Yuyao’s own automated laboratory provides model-driven validation around the clock. The platform calls specialized sub-models to propose candidate compositions, crystal configurations and process recipes; an automated line carries out synthesis; XRD phase characterization and SEM/EDS morphology and composition analysis produce measured data; and every measurement is compared against the model’s prediction and fed back for retraining. Each round of validation is what drives the model’s next improvement.

MatterSeek is already deployed in embodied-intelligence robotics, semiconductors, new energy and biodegradable materials. It serves dozens of industrial customers and has completed more than one hundred R&D assignments to date.

Users can reach the platform through a web browser or a downloadable client. For enterprise customers, Yuyao offers private deployment and data residency options, with intellectual property in sensitive formulations remaining entirely under the customer’s control to meet confidentiality and compliance requirements.

“MatterSeek does not deliver a static report. It delivers a technical plan that can be put into practice.”

That is what separates MatterSeek from other materials AI platforms, and where the value of the closed-loop paradigm lies. Industrial usability is the benchmark: full-dimensional performance modeling, multiphysics validation, process plans that can be executed and a closed experimental loop combine to turn an industrial requirement into a verified set of material candidates and a manufacturing-ready process.

MatterSeek is AI materials software: a platform that designs a material against a performance target, instead of testing candidates one at a time. What distinguishes it within the category is where it stops. Most tools return a predicted structure. MatterSeek returns the structure and the way to make it — composition, crystal lattice, processing route, required equipment and an estimate of raw-material cost.

Materials development is slow because of the loop, not the idea. A single cycle — propose, synthesize, characterize, adjust — takes months, and most candidates that enter it fail. MatterSeek compresses both ends of that loop. Inverse design generates candidates from the target rather than from intuition, so the search begins closer to a viable answer. The platform then returns a synthesis route with a cost estimate, so the first experiment is one worth running.

Accuracy is settled before anything is made. High-value candidates are checked with DFT, phase-field and finite-element simulation, and any of the 190 material metrics can be weighted as an objective or fixed as a constraint. Confirmation happens in Yuyao’s own automated laboratory, which runs model-driven experiments around the clock, characterizes each result by XRD and SEM/EDS, and returns the measurement for retraining. Every cycle therefore improves the next one instead of merely consuming it.

The effect is a shorter path from specification to a manufacturable material: fewer design–make–test cycles, and a cost figure early enough to change the design.

International website: https://www.yuyaomatterseek.net 

Web access and client download, with access to Yuyao’s in-house laboratory for experimental validation.

China beta site: https://yuyaomatterseek.com 

Free online trial open to users in China.

E-mail: matterseek@gmail.com