Skip to main content
The European High Performance Computing Joint Undertaking (EuroHPC JU)

Foundation models for atomistic simulations via Active Learning

50000
Awarded Resources (in node hours)
Leonardo BOOSTER
System Partition
January 2025 - January 2026
Allocation Period

AI Technology: Deep Learning

This project combines the state-of-the-art data mining technique called Active Learning with the recently developed FeNNol library for training Machine-Learning-based force fields. 

With the help of the Active Learning and the EuroHPC resources, the team intends to augment existing datasets in order to train and validate the new generation of reference foundation models for biological and chemical applications.

The models provided by FeNNol are accurate, transferrable and scalable, which makes them promising candidates for impactful applications in molecular dynamics, drug discovery, spectroscopy and catalysis.