Dutch AI company designs new polymers to specification, in weeks instead of years

TNO today launches polySCOUT, an independent Dutch AI company that designs polymers to specification. Target properties and regulatory constraints go in; lab-validated candidate materials come out. Where the design and selection phase of a new polymer typically takes years, polySCOUT brings it down to weeks. The technology has been developed within TNO since 2022, and TNO becomes a shareholder in the new company. To mark the launch, the founders ring the gong at Euronext Amsterdam this morning.

Why now

Plastics are everywhere: 431 million tonnes were produced worldwide in 2024 (PlasticsEurope). Yet the molecules the industry relies on are largely a century old, and regulation is now changing faster than the chemistry. Since 20 July 2026, bisphenol A has been banned in food contact materials in the EU. The European microplastics restriction phases out synthetic polymer microparticles category by category through 2035. A broad PFAS restriction is in preparation. Every deadline means the same thing: a material has to go, and its replacement has to perform at least as well.

"Forced substitution is a performance problem with a deadline," says Michaël van der Jagt, co-founder and CEO of polySCOUT. "Conventional polymer development succeeds in five to ten percent of cases. Not because chemists lack skill, but because the search space is unimaginably large: more than 10⁶⁰ possible polymer structures. We don't make R&D faster. We make it start in the right place."

How it works

polySCOUT turns the classical development process around. Instead of making a molecule and then measuring what it can do, the customer specifies what the material must do: for example a glass transition temperature above 105 degrees, clarity, dishwasher resistance and no bisphenol. polySCOUT's AI models are trained on a proprietary, verified database with thousands of data points on polymer properties, built up over four years of research and manually checked back to the source. With it, the models search tens of thousands of possible copolymer combinations of industrially available, REACH-registered building blocks for materials that meet the specification, weighing safety, cost and end of life along the way. The most promising candidates are synthesised and measured in a high-throughput laboratory: 25 physical materials per week. The customer receives not a prediction, but a measured material. Because polySCOUT can select building blocks that are already REACH-registered and available at industrial scale, a candidate it finds is directly scalable. No exotic lab chemistry, but materials a producer can make in the near term.

Polymer-native, and deliberately not broad

In a year in which generalist AI materials companies in the United States and the United Kingdom raised hundreds of millions of dollars to search "all materials", polySCOUT deliberately chooses depth. The company does one thing: polymers.

"A polymer is not a molecule but a distribution: chain length, sequence, branching, crystallinity," says Milad Golkaram, technical co-founder of polySCOUT. "Models built for crystals or small molecules don't see that. We built our platform for polymers from the ground up: a polymer-native representation, a proprietary verified database of polymer properties, and models trained on what polymers actually do. That is not a detail. That is an architectural difference."

From TNO programme to company

polySCOUT started in 2022 as a TNO research programme under the banner "sustainable polymer design by machine learning". Four years on, there is a verified database, validated predictive models and a working chain from data to physical material. polySCOUT B.V. takes over the technology under an exclusive licence from TNO; TNO becomes a shareholder. The company is currently in discussions with product manufacturers and material producers about its first design programmes, and this autumn opens a first external financing round to scale the platform and the team.

"The materials Europe will run on in 2050 are being designed now. Whoever masters that design process determines part of their own industrial future. With polySCOUT, we are putting Dutch knowledge to work to build that position," says Tjark Tjin-A-Tsoi, CEO of TNO.

polySCOUT is led by three co-founders: Michaël van der Jagt (CEO), serial deeptech entrepreneur with more than twenty years of experience in polymer materials and international manufacturing; Ellen de Ruiter (CCO), chemical engineer and former McKinsey consultant in chemicals, plastics and sustainability strategy, who has shaped polySCOUT's commercial development from within TNO since 2023; and Milad Golkaram (CSO), polymer scientist with a PhD from the University of Groningen, who was present at the inception of the polySCOUT programme at TNO.

About polySCOUT

polySCOUT is a Dutch AI company that designs polymers to specification. Its platform combines polymer-native AI models, a proprietary verified database with hundreds of thousands of data points on polymer properties, and high-throughput lab validation into a single chain from specification to measured material. Material producers and brand owners receive validated candidates for substitution and performance challenges within weeks, from bisphenol-free copolyesters to heat-resistant polyamides. polySCOUT grew out of four years of research at TNO and has been an independent company since August 2026, with TNO as a shareholder.