materials-project

A materials discovery algorithm geared towards exploring high-performance candidates in new chemical spaces.

Use time-splits for Materials Project entries for generative modeling benchmarking.

[npj Computational Materials] The implementation for the paper "Physics-informed graph neural network representation learning for crystal property prediction"

Interactive phase diagram generator for up to 4 components (solid phases) using the Materials Project API and pymatgen.

AI-native materials discovery workspace with graph-grounded evidence and agentic candidate screening

Automatic Generation of Quantum ESPRESSO Input Files

Modern React UI library based on mp-react-components, for materials science applications, with full TypeScript support and reusable scientific UI components.

Transfer learning using a high order message passing foundational model to F-Li-B cells.

Some programs I use in my research

Scientific data enrichment tool for Open WebUI - Chemistry and materials science integration with PubChem, ChEMBL, Materials Project, and RDKit

Predicting lithium cathode properties (voltage, capacity, formation energy, stability) from crystal structure using CGCNN, M3GNet, TensorNet, Random Forest, and XGBoost on data from Materials Project, OQMD, AFLOW, and JARVIS

Deep Learning toolkit for Materials Science. Predict properties (CGCNN, MEGNet, M3GNet), discover new materials (GNoME-inspired), simulate XRD, and serve via GraphQL API.

Independent reproduction of ECSG thermodynamic stability prediction with fixed Materials Project split, released checkpoints, and paper-vs-reproduction analysis.

Does crystal symmetry predict oxide stability? Four-part analysis on Materials Project data: exception detection, composite scoring, polymorph case studies, and solar band gap screening.

Provenance-first phase scouting and indexed powder diffraction references

Criticality-aware sodium-ion cathode prioritization using Materials Project records, text-mined evidence, and leakage-audited ML.

ML pipeline that screens the Materials Project for overlooked solar absorbers — surfaced a validated 1.3 eV candidate with zero PV literature

Multimodal Deep Learning pipeline for crystalline bandgap prediction. Combines 1D X-Ray Diffraction (XRD) patterns with engineered tabular features (Magpie & CrystalNN) using a dual-branch PyTorch ResNet architecture. Features high-throughput data extraction via Materials Project API and automated structural featurization.

Side quests: parametric CAD as code, espresso machine telemetry, medical-imaging pipelines, and local-first AI tooling - built evenings and weekends with honest engineering notes