Eindhoven University of Technology (TU/e) will receive 1.5 million euros within the European Horizon Europe project SimuLingua, an initiative with a total budget of 6 million euros that aims to develop a next-generation AI project for materials discovery. TU/e is the largest academic partner within the consortium, with 25 per cent of the total budget, and is using the funding to appoint five postdoctoral researchers in Eindhoven.
The SimuLingua project, led by Flowphys AS (Norway), brings together partners from the Netherlands, Sweden, the UK, Ireland, Lithuania and Ukraine. In a highly competitive European call, the project finished second out of 47 proposals submitted, with a near-perfect score of 14 out of 15. This high rating underlines SimuLingua's potential to radically change the way materials are designed and developed in Europe.
Multimodal model
SimuLingua aims to fundamentally innovate the development of materials by engineers and scientists with an AI system that seamlessly switches between natural language, physics simulations, images and data from experiments. Within the AI project, an open, multimodal 'scientific foundation model' is being developed that supports a closed design loop that goes from design, to simulation to validation.
Exploring new materials
This will allow researchers to explore the potential of new materials through natural language interaction, with underlying AI models verifying whether proposed designs are physically feasible. The platform should significantly speed up the discovery of advanced materials by allowing experts to quickly and efficiently test ideas against reality using computer simulations.
Two developments thanks to AI project
"This project brings together two ground-breaking developments: AI foundation models and technical simulations based on fundamental laws of nature" says Victorita Dolean, professor at TU/e and the university's contact person for the project. SimuLingua allows researchers to 'talk' to material models, so to speak, explore ideas interactively and test them quickly in silico (via computer simulations rather than in a physical laboratory). For TU/e, this represents an important step towards AI-driven engineering sciences.
This project brings together two pioneering developments: AI foundation models and engineering simulations based on fundamental laws of natureVictorita Dolean, professor at TU/e
Software development
SimuLingua's impact extends beyond academia. An industrial partner within the consortium points out that AI tools have already dramatically changed software development and expects SimuLingua to have a similarly disruptive effect on engineering and science.
European blueprint
The AI project aims to establish a European blueprint for scientific foundation models, opening the way to more efficient and scalable materials research in a variety of sectors. A foundation model is a large AI model that can serve as the basis for a variety of different applications.
From fundamental research to practical applications
The project runs for 48 months and aims to take the technology from TRL-1, the basic research phase, to TRL-4: a laboratory-tested and validated prototype. This schedule shows that SimuLingua not only wants to develop an AI model, but also to demonstrate that it is actually useful for material discovery.
Strengthening the regional AI ecosystem
For TU/e, the award means a strengthening of the university's position in AI, computational science and engineering. Prof Victorita Dolean, will search for candidates to fill five postdoc positions in Eindhoven, assisted by Wil Schilders, Björn Baumeier, Lisa Kusch and Michael Abdelmalik as core team members.
AI project and other initiatives
The researchers focus on machine learning for physics, high-performance computing and trusted AI for critical engineering applications. The project thus fits well with other initiatives in the Brainport high-tech ecosystem and strengthens the region's role as a centre for advanced technology and innovation.
Source: TU/e, Martijn Luyk; Photo by Gerd Altmann via Pixabay
Read also: Quickly and automatically identifying welding defects with artificial intelligence
