Skip to main content

New Finnish LUMI projects chosen: creating fully open multilingual language models, designing vaccine adjuvants and more

LUMI data center

The third batch of Finnish research projects to utilize LUMI, Europe’s fastest supercomputer, have been chosen, and will run on the system using its massive GPU capacity. The projects are, for example, creating Large Language Models (LLMs) for all official European languages, developing novel optimized vaccine formulations and more.

The resources were granted in the Third Finnish LUMI Extreme Scale Call, which shared resources from Finland’s country share of the LUMI resources.

The accepted projects are:

• CHIRAL-SENSE: Designing Biosensors from Chiral Gold Nanoclusters, PI Hannu Häkkinen, University of Jyväskylä
The project aims to investigate the interactions between chiral gold nanoclusters and chiral biomolecules in water by means of massive parallel molecular dynamics simulations and electronic structure calculations. The study focuses on understanding the contribution of conformational properties (chirality and enantiomerism) towards molecular selectivity in biological systems which has been suggested to have possible applications in design of highly-selective drug molecules.

• LumiLLM: LUMI Large Language Models , PI Sampo Pyysalo, University of Turku
The LumiLLM project will create the largest fully open, massively multilingual language models (LMs) with full coverage of all official European languages, creating the basis for the next generation of multilingual generative AI systems, and enhancing the digital sovereignty of Europe and countries beyond its borders. The project will leverage the vast computational capacity of LUMI and the extensive text resources of the Horizon Europe project HPLT (High Performance Language Technologies) to train LMs of up to 100 billion parameters on trillions of words from diverse sources, establishing a new state-of-the-art for open models. The resulting models will enable a broad range of advances in AI across dozens of languages, creating new opportunities in research and industry and taking steps towards more generally intelligent systems. By making all of its tools and models as well as their detailed training process openly available, the project will also contribute toward transparent and trustworthy European AI. This project will be carried out in collaboration with Silo AI, one of Europe’s largest private AI labs. You can read more about this project in Silo AI’s press release.

• Pre-Exa-vaccine: Pre-Exascale computing as a design tool for vaccine adjuvants and delivery vehicles, PI Alex Bunker, University of Helsinki
The project aims to carry out large-scale all-atom molecular dynamics simulations for investigating the effect of two vaccine excipients (polymer sheath of lipid nanoparticles and adjuvants) on modulating the surface properties of drug delivery systems. Findings from the study have been proposed to provide necessary mechanistic insight needed for the development of novel optimized vaccine formulations that enhances targeted drug delivery while alleviating side effects associated with currently used vaccine formulations.

The next Finnish LUMI Extreme Scale Call will open in September 2023.

More information about applying for LUMI’s resources on the Get started section.