Alloy & Materials Design · 03
LeilMat
AI-native canvas for alloy design. Periodic-table spatial editor, validated property prediction, and click-to-trace provenance on every number.
How it flows
What lives inside
Property Prediction
Mechanical and thermal properties from composition — RF + physics features, family-specific calibration, temper-aware encoding.
Inverse Design
Target a property envelope, get Pareto-optimal alloy candidates. Engineering briefs explain the why for each suggestion.
Process Simulation
CALPHAD equilibrium · Scheil solidification · Fe-C TTT · microstructure · KWN precipitation with aging-playback scrubber · 1-D Fick diffusion · Neper grain structure · Potts grain growth · DAMASK CPFEM stress–strain · heat-treatment aging optimizer.
Agentic AI Copilot
The AI doesn't just narrate — it emits clickable action pills that mutate the canvas. "Set Mg to 4 %" / "Cap density at 2.7" / "Run Predict" — design steps you accept with one tap.
Audit-grade Reports
Every design exports as a multi-page PDF: composition, property table with ±CI, radar chart, phase + KWN trajectories, AI rationale, full provenance bibliography. Ready for a design-review packet.
Conformal Uncertainty
Every ML prediction carries a calibrated 95 % confidence interval plus an out-of-domain flag. No black-box numbers — provenance and uncertainty travel with the value into the report.
Built on a broad scientific harness — pycalphad equilibrium, Scheil-Gulliver + Fe-C TTT, KWN precipitation kinetics with an interactive aging-playback scrubber, a 1-D Fick diffusion mini-sim, a Taichi-JIT anisotropic phase-field dendrite simulator, Neper grain structures with Potts grain growth, DAMASK crystal-plasticity (CPFEM) stress–strain, a heat-treatment aging optimizer, and conformal-calibrated ML models for mechanical properties. An agentic AI copilot proposes concrete design steps you apply with one click; the canvas exports as an audit-grade PDF report. See how it works →