AI Cathode Screening

Accelerate Battery Material
Discovery with AI

Instantly predict thermodynamic stability of Li-cathode materials using our governance-approved MACE-MP-0 ensemble. 92.7% KEEP precision, 0% false-kill rate, 90% calibrated coverage.

1

Upload Structure

Drag & drop your CIF file. Our model parses 3D crystal structures instantly.

2

MACE Ensemble Inference

5-member MACE-MP-0 fine-tuned ensemble predicts Ehull with conformal calibration.

3

Get Recommendation

Receive KEEP, MAYBE, or KILL recommendations to prioritize your research.

Built on Rigorous Science

We leverage a 5-member deep ensemble of MACE-MP-0 (Multi-Atomic Cluster Expansion) foundation models, fine-tuned on 17,227 Li-cathode materials from the Materials Project.

  • SOAP-LOCO Validation

    Validated on clustered hold-out sets to ensure generalization to new chemistries.

  • Uncertainty Quantification

    Aleatoric (quantile regression) + epistemic (inter-model disagreement) + conformal calibration for 90% coverage.

  • Decision-Grade Output

    Clear KEEP/MAYBE/KILL recommendations for immediate lab prioritization.

Read full methodology

2.12×

Enrichment Factor

EF@10 on test set

0.030

MAE (eV/atom)

Test set (1,013 materials)

0.0%

False Kill Rate

No stable materials discarded

17,227

TMO Cathode Materials