Software & code
Open-source tools, pipelines & datasets
All code and datasets associated with publications are made publicly available on GitHub and Zenodo upon publication. Below are the main repositories.
RuP phase transition analysis pipeline
End-to-end workflow for fine-tuning MACE-MH-1 on DFT data, running multi-GPU LAMMPS molecular dynamics on LUMI and Mahti, and performing symmetry, RDF, structure factor, and lattice parameter analysis.
Associated with: ChemRxiv 2026 · Repository and Zenodo deposit made public upon journal publication.
Python
MACE
LAMMPS
VASP
¹²⁹Xe NMR ML pipeline for porous liquids
Equivariant GNN training pipeline for ¹²⁹Xe chemical shift prediction in CC3 porous organic cages. Includes data generation scripts, model training, and MD-coupled shielding computation.
Associated with: J. Phys. Chem. Lett. 2025 ·
GitHub: ozakary ·
Zenodo
Python
PyTorch
e3nn
ASE
Multi-element ML NMR models
Benchmark suite for kernel ridge regression, SchNet, and MatTen NMR shielding models across ¹H, ¹³C, and ¹⁵N, with relativistic DFT training data generation via ORCA and VASP.
Associated with: J. Phys. Chem. A 2026 ·
GitHub: ozakary
Python
SchNet
scikit-learn
ORCA