Every major industry built a digital brain.
Metals never did. We're changing that.
METALLAI turns weeks of metallurgical trial-and-error into overnight answers. Physics-aware AI for alloy prediction and inverse design — so engineers stop guessing and start designing, on the first try.
to hit target properties
with quantum-hybrid AutoForge
validated 42CrMo4 case
& rework globally
Pharma has computational discovery. Semiconductors have process simulation.
Metals still run on trial & error.
Every alloy, every heat, every process window — still figured out one melt at a time. Decades of metallurgical intuition are retiring out of the industry with nowhere to go.
— The Cost of Trial & Error
— What METALLAI Changes
- Multiple trials to hit spec
- Each off-spec heat → scrapped
- Weeks of delay per order
- Relies on senior metallurgist intuition — no digital record
- 1–2 high-confidence trials
- Off-spec heats rescued by adjusting heat treatment
- Predictions in seconds
- Expertise captured as computable knowledge
Five modules. One intelligence layer.
Prediction, fatigue analysis, inverse design, welding consumable selection, and an AI assistant that chains them together — built on the same physics-aware backbone, deployed end-to-end in the browser.
Mechanical Predictor
Composition + process + heat treatment → full mechanical property profile with calibrated uncertainty bands. Hybrid ML + physics, not a black box.
- YS, UTS, hardness (HV), elongation with 95% CI
- Hall-Petch grain-boundary strengthening decomposition
- Zener-Hollomon hot-working physics (forging / rolling)
- Chvorinov / Niyama / SDAS for casting routes
- Process feasibility gate (Ac1/Ac3, solidus/liquidus)
- Domain-distance trust scoring (Mahalanobis)
Fatigue Predictor
High-cycle and low-cycle fatigue life from composition + heat-treatment + stress state. Defect-sensitive, surface-condition-aware, R-ratio aware.
- S-N curves & endurance limits with confidence bands
- Basquin (HCF) + Coffin-Manson (LCF) HT-aware coefficients
- Murakami √area for defect-driven cases (castings)
- Goodman / Gerber mean-stress correction
- Paris LEFM for crack growth screening
- Carries seamlessly from mechanical predictions
Inverse Alloy Design
Flip the question. Give AutoForge target properties — it returns feasible composition + process recipes, ranked by trade-offs you choose.
- Differential Evolution + Bayesian Optimization + quantum-hybrid search
- Pareto-front trade-off landscape (cost / manufacturability / properties)
- Fatigue & weldability gated into the search loop
- Process-route aware: forging vs casting vs rolling
- 17× faster design search vs. classical evolutionary methods
- Auto-generated certification-ready PDF reports
Welding Consumable & HAZ Predictor
Ranked consumable recommendations + predicted weld-zone properties + defect-mode probabilities. Built for offshore, Arctic, and sour service.
- Ranked filler-metal candidates (AWS A5.x classification)
- Weld-zone YS / UTS / HV / EL / CVN with p10/p90 bands
- CVN target temperatures from +20°C (room) to −100°C (LNG cryogenic)
- Defect probabilities: porosity, HAC, sol-cracking, reheat cracking, lamellar tearing, LoF
- PWHT suggestions tied to alloy + service environment
- Carries from multilayer base-metal predictions (full audit chain)
Conversational Metallurgy Copilot
Talk to your alloy data in plain English. METALL runs predictions, chains multi-step analyses, and explains the physics behind every answer.
- Runs Mechanical / Fatigue / Multilayer / AutoForge / AutoWelding on command
- Chains workflows: mechanical → multilayer → welding in one conversation
- Carry-from-prior-prediction: no re-typing composition
- Explains the metallurgy behind every prediction (Hollomon-Jaffe, Koistinen-Marburger, etc.)
- User-confirmed write actions (you approve every job before it runs)
- Two tiers: Basic (fast) and High (deep reasoning)
Give us the targets.
Get back a family of feasible routes.
Engineers don't ask "what are the properties?" — they ask "how do I achieve them?" AutoForge inverts the problem: from target window to composition + process, in seconds.
Define Targets
Set yield, UTS, elongation, hardness — plus composition & process constraints.
Virtual Exploration
Quantum-hybrid search scans millions of composition + process combinations autonomously.
Physics Validation
Every candidate gated by thermodynamic, kinetic & weldability checks. No hallucinations.
Pareto-Ranked Results
A family of feasible routes — balanced, aggressive, cost-optimal. You pick the trade-off.
From our LinkedIn series:
real alloys, real labs, real numbers.
Two case studies showing METALLAI in action — one for forward prediction, one for inverse design. Both experimentally validated or benchmarked against industry practice.
Can AI Reduce Alloy Development to a Single Trial?
Target window: 285–340 HB hardness, 880–1080 MPa UTS. Traditional approach would need 5–10 iterations.
Same Alloy. Same Composition. Different Performance.
Targets: 1150 MPa YS, 1300 MPa UTS, 13% elongation. AutoForge returned two legitimate process routes — balanced vs. aggressive.
Strength. Manufacturability. Reliability.
All on the table — at the same time.
METALLAI doesn't just predict numbers. It surfaces the three-way trade-off that every metallurgist knows exists but rarely sees quantified.
Fewer experiments
To hit a target property window — validated on 42CrMo4 and benchmarked on 300M aerospace steel.
Faster design search
Quantum-hybrid annealing reaches the same Pareto front in 92 s vs. 1,611 s classically.
Prediction accuracy
Validated on 100+ data points at a major Turkish metal manufacturer's R&D facility.
Can we design alloys that don't exist yet?
Not magic — the same physics-aware search, pointed at a blank slate instead of a known grade.
Predict & Optimize
Hit a target window inside a known alloy family. Composition + process → properties.
Inverse Design
Feasible composition + process routes — with fatigue, weldability & uncertainty gated in.
Novel Alloys
The inverse engine proposes compositions outside the catalogue — from targets alone.
From prediction to design — bridging physics, data, and decisions.
Backed by & built with
See METALLAI on your own alloy.
Launch the browser predictor, or request a tailored demo with our team.
