진학프로석·박사 채용 접수 플랫폼

Are you excited about using large-scale AI to accelerate scientific discovery? Join a Horizon Europe project developing next-generation scientific foundation models that combine knowledge graphs, multi-modal data, and GPU-accelerated machine learning for materials science.

Information
We are seeking two highly motivated postdoctoral researchers to join the Horizon Europe project SIMU-LINGUA, a major European initiative developing scientific foundation models (SciFMs) for materials science.

SciFMs are emerging as a powerful paradigm for scientific discovery. SIMU-LINGUA addresses key challenges in data integration, model design, and large-scale training by combining multi-modal scientific data, knowledge graphs, physics-aware machine learning, and GPU/HPC computing to develop transparent and trustworthy AI for science.

At Eindhoven University of Technology (TU/e), you will contribute to the project’s core technical components:

• • scientific data orchestration and knowledge graphs
• • architecture and large-scale training of scientific foundation models

You will collaborate with researchers across machine learning, scientific computing, materials science, and data engineering, and work with leading academic and industrial partners across Europe. The project will develop materials ontologies, training-ready datasets, multi-modal knowledge graphs, and large-scale SciFM models, together with tools for training diagnostics and model observability. The first pre-trained SciFM models will be released as part of the project. You will have significant scientific ownership, contribute to publications and open-source software, and help shape emerging methodologies for scientific AI and foundation models. Applicants should indicate a primary research track, although collaboration between tracks is expected.

Track A — Scientific Data & Knowledge Graphs
You will develop scalable scientific data infrastructures enabling large-scale model training, including materials ontologies, data ingestion and curation pipelines, multi-modal knowledge graphs, and training-ready datasets with robust provenance and validation.

Track B — GPU-Accelerated Scientific Foundation Models
You will design and train large-scale multi-modal foundation models, including SciFM architectures coupled to knowledge graphs, GPU-accelerated PyTorch training pipelines, distributed training on HPC systems, and tools for training diagnostics and observability, potentially integrating physics-aware constraints and generative modelling approaches.

Both tracks interact closely to create a data–model feedback loop, enabling systematic analysis of how scientific data, model architectures, and training dynamics influence scientific predictions. You will join a vibrant research environment at TU/e at the intersection of AI, scientific computing, and computational science, collaborating with leading European research groups and benefiting from advanced GPU and HPC infrastructure.