Skip to content
HolonomiX
Products/HX-MDAO-TP
Deployed · v1.0

HX-MDAO-TP

Validity-gated, reproducible, signed — or no number at all.

A trust and provenance plane woven into the type system: every result provably inside a validated region, or explicitly marked outside it. Never a confident wrong answer.

Statusv1.0 — gate-verified
ValidationAcceptance suite · CPU + A100
Formal proofLean4 kernel-checked certificate
SigningML-DSA at commit boundaries
ProcurementDirect, incl. government channels
Couple Surrogate Optimize Propagate Stamp Sign
Product surface

The working console.

Model, Operate, Prove, Compare. The console renders the engineering system without interpretation drift, and treats the evidence package — not the rendered result — as the claim. Trust states are engine-issued; the browser cannot upgrade them.

Model · structure + semantics
HX-MDAO-TP Model view: discipline graph with coupling topology, contracts, representations, and the hashes that bind them.
Prove · evidence surface
HX-MDAO-TP Prove view: engine-issued verdict with validity policy, evidence refs, ledger chain, release binding, and ML-DSA-65 signer identity.

The multidisciplinary design engine you can put in front of a certifier.

HX-MDAO-TP is a universal engine for multidisciplinary design analysis and optimization with a trust and provenance plane woven into its type system. It couples your disciplines, builds representation-aware response surfaces across topology changes and tested validity regions, drives the optimizer, propagates uncertainty, and returns a result stamped with its validity domain and sealed with a post-quantum signature.

Most optimization stacks give you an answer. HX-MDAO-TP gives you an answer you can defend. Every number it hands back is either provably inside a validated region or explicitly marked as not.

Two independent levers. No losing category.

The compression lever

Problems that are infeasible dense become tractable.

A self-authored tensor-train / QTT substrate represents fields, operators, and Jacobians in compressed form and differentiates through them, delivering bounded-rank surrogates for high-interaction-order quantities where the tested spectral baselines produced no qualifying result.

The trust lever

An answer that is validity-gated, reproducible, and signed.

Best-of-breed numerics plus a provable, attested trust and provenance plane: results signed with ML-DSA at commit boundaries, recorded in an append-only chained ledger, and verifiable by a third party from the file alone.

The honesty invariant

A result is IN_VALIDATED_DOMAIN only inside the anchors you supplied. Outside them it is returned and explicitly marked OUT_OF_VALIDATED_DOMAIN. Where the model did not converge, you get no number at all rather than a confident wrong one.

v1.0

Gate-verified end to end: a full acceptance suite passing on both CPU and NVIDIA A100, dense-CPU-versus-A100 agreement to floating-point round-off, a kernel-checked Lean4 certificate, and a signed, third-party-reproducible provenance manifest over the shipped tree. Deploys as a hardened, reproducible image and runs offline with no network attack surface.

Optimize what was infeasible. Sign every number you ship.