Community-driven, well-tested, and reproducible scientific tools developed by the Data Driven Materials and Molecular Science group.
🪐 janus-core
Python / ASE / CLI
Tools for materials modeling with machine-learned interatomic potentials (MACE, SevenNet, CHGNet, M3GNet). Provides a simple, robust CLI and Python API with full ASE calculator support.
pip install janus-core
Single-point energies, forces, stress tensors, and Hessians
Geometry optimization with FrechetCellFilter and BFGS/FIRE
Molecular dynamics in NVE, NVT, and NPT ensembles with thermostat/barostat logging
Automated equation of state (EOS) and full 6x6 elasticity stiffness tensors ($C_{ij}$)
Phonon band structures & DOS via Phonopy integration
An open-source AiiDA plugin integrating the janus-core library to manage automated workflows for machine learning interatomic potentials (MLIPs) with complete data provenance.
pip install aiida-mlip
Full provenance graphs recording every calculation input, output, and potential parameter
Automated single-point, geometry optimization, and molecular dynamics workchains
Scalable execution across local workstations and remote HPC clusters
Integrates directly with the wider AiiDA simulation and materials informatics ecosystem
An interactive browser-based AiiDAlab application for configuring and running machine learning interatomic potential calculations with AiiDA and aiida-mlip.
pip install aiidalab-mlip
Interactive web interface for submitting MLIP simulations without writing boilerplate code
Structure loading from CIF, XYZ, and standard crystallography formats with 3D visualization
Direct integration with pre-trained foundation models (MACE-MP and others)
Interactive single-point energy, force evaluations, and geometry optimizations with full AiiDA provenance
A Python package and CLI for building realistic atomistic and molecular systems for materials modeling, utilizing machine-learned interatomic potentials (via janus-core) with Monte Carlo and Molecular Dynamics routines.
pip install pack-mm
Generates realistic packed starting configurations for complex liquids, interfaces, and porous frameworks
Uses janus-core for MLIP interactions, with MACE-MP foundation models enabled by default
High-performance packing leveraging Monte Carlo, Molecular Dynamics, and hybrid MC/MD relaxation
Full Python API and intuitive CLI with support for both CPU and CUDA GPU acceleration
Machine Learning Performance and Extrapolation Guide. A comprehensive benchmarking framework and interactive performance guide to evaluate MLIPs across diverse chemical systems, extrapolations, and physical observables.
git clone https://github.com/ddmms/ml-peg.git
Evaluates model performance beyond basic force/energy errors to actual physical stability
Tests out-of-distribution generalization, extrapolation limits, and uncertainty quantification
Interactive web dashboard for comparing foundation models and dataset baselines
Open-source protocols for standardized community potential verification
A web application and library for automated generation of input files with optimised k-point meshes for Quantum ESPRESSO self-consistent field (SCF) calculations. Developed as part of the PSDI (Physical Sciences Data Infrastructure) Data to Knowledge initiative to eliminate computational waste and improve sustainability on national supercomputers like ARCHER2.
pip install goldilocks
Part of the UKRI PSDI Data to Knowledge national infrastructure framework
Predicts "Goldilocks" k-point convergence parameters to reduce compute waste and carbon footprint