WHAT WE BUILD
Our Work
From open materials datasets to physics-aware foundation models, we build the infrastructure that enables researchers to accelerate materials discovery with AI.
A harmonized dataset combining Materials Project, OQMD, and JARVIS with quality scores, provenance tracking, and standardized material properties for AI-driven materials research.
- I.267,230 materials from 3 databases
- II.Quality-scored and validated
- III.Cross-source provenance tracking
- IV.CC BY 4.0 open license
Physics-aware graph neural networks trained on large-scale crystal datasets for transferable materials property prediction across diverse material families.
- I.Physics-informed modeling
- II.Multi-property prediction
- III.Transferable across families
- IV.Benchmarked against state-of-the-art
Accelerating screening of solid electrolytes, cathodes, and anodes through AI-guided discovery pipelines purpose-built for solid-state battery materials.
- I.Solid electrolyte screening
- II.Cathode discovery
- III.Ionic conductivity prediction
- IV.Stability against lithium metal
Open datasets, reproducible benchmarks, and developer tools that make AI-driven materials research accessible to the global scientific community.
- I.Open datasets on Hugging Face
- II.Reproducible benchmarks
- III.Model cards and documentation
- IV.GitHub open-source tools
SCANDIUM DATASET
