Accelerating Materials Science through AI

Accurate Prediction Models,Robust Software Frameworks,Curated Materials Datasets


Web apps implementing pre-trained property prediction models.


MatErials Graph Network (MEGNet) models for formation energies, band gaps, elastic constants and other properties of crystals.


Deep neural networks for formation energy prediction of garnets and perovskites.


Random forest models for prediction of coordination environment from X-ray absorption spectra.


Software frameworks for robust AI in materials science.


Library for building graph networks for materials science developed by the Materials Virtual Lab.


Library for machine learning for materials science developed by the Materials Virtual Lab.


Data mining tool for materials science developed by the Hacking Materials group.


Implementation of the Crystal Graph Convolutional Neural Network (CGCNN) for prediction of crystal properties.


The Atomic Energy NETwork (AENet) package is a collection of tools for the construction and application of atomic interaction potentials based on artificial neural networks (ANN) developed by


The Global Attention Graph Neural Network (GATGNN) package implements a graph neural network for improved inorganic materials' property prediction.


Curated datasets for reproducible AI in materials science.

MP 2019.4.1

Graphs and formation energies of ~133,000 Materials Project Crystals as of Apr 1 2019.

MP 2018.6.1

Graphs, formation energies, band gaps and elastic constants of ~60,000 Materials Project Crystals as of Jun 1 2018.

QM9 Molecules

Graphs of the QM9 set of ~134,000 molecules developed by Ramakrishnan et al.


Datasets for the development of Spectral Neighbor Analysis Potentials for various elements, alloys and compounds.

Garnet and Perovskite

Dataset for the development of neural networks that predict stability of garnets and perovskites. Model files included.