New success: Project NoE-AIMat

The project:

NoE-AIMat establishes the first European Network of Excellence for AI in materials science as a RAISE pilot under HORIZON-RAISE-2026-01-01, uniting 22 partners from 14 countries over 48 months. Despite over 500,000 annual materials-related publications, AI capabilities remain fragmented across incompatible datasets and disconnected workflows. Frontier LLMs hallucinate at the research frontier; full PSPPA chain extraction remains unsolved; SSbD is treated as post hoc; and no coordinated European mechanism sustains a shared AI-for-materials ecosystem.

NoE-AIMat addresses these gaps through five advances:

(1) federating The World Avatar (TWA) and OpenSemanticLab (OSL) via the OO-LD schema into a unified AI operational knowledge infrastructure;

(2) introducing Materia (Materials Innovation Agent), combining RAG from federated knowledge graphs, LoRA/QLoRA adapters, and Model Context Protocol (MCP) tool orchestration across European HPC and AI Factory resources (CSC/LUMI, CINECA);

(3) pioneering scalable PSPPA chain extraction from literature into machine-readable causal knowledge structures;

(4) operationalising SSbD as a concurrent multi-objective design constraint via openLCA integration in Materia; and

(5) building a self-sustaining RAISE ecosystem with a living

SRIA, shared benchmarks, structured talent exchange (120 mobility person-months), and three FSTP calls engaging 28 external beneficiaries.

Research targets grand challenges in inverse molecular design, multi-scale fatigue prediction, and SSbD materials discovery. By M48, the project delivers 5 million knowledge graph entities, 120 datasets, 100 MCP tool integrations, 60 publications, and 20-fold workflow acceleration, securing European leadership in AI-driven materials science.