Platform Architecture

The DUNS Number for Art

MIRAS is not a tool. It is the identity and compliance infrastructure for the global art market — an open platform where every artwork has a verifiable, permanent, computational identity.

Confidential — MARGIN International LLC · July 2026

The gap

Every industry has a universal identity system.
Art does not.

Companies have DUNS. Securities have ISIN. Books have ISBN. Products have UPC. Finance has Bloomberg. None of these existed until someone built the infrastructure and made it the standard. Art is the last major asset class without a computational identity layer.

🏢

Companies

DUNS Number — universal business identifier. 500M+ records. Required for government contracts, credit checks, supply chain verification. Built by Dun & Bradstreet starting 1963.

📊

Finance

Bloomberg Terminal — universal financial data layer. 325,000+ subscribers. $12B+ annual revenue. Every institution relies on it because everyone else does. Network effect is the moat.

🎨

Art

Nothing — no universal identifier. No computational proof of identity. No standardized risk scoring. Provenance lives in paper files, expert opinions, and institutional memory. This is a $65B+ market operating on trust.

The analogy

MIRAS fills the same structural gap for art
that these systems filled for their markets.

Companies

DUNS

Universal business ID. 500M+ entities. Required for contracts and credit.

Securities

ISIN / CUSIP

Standard identifier for every traded security. Settlement requires it.

Finance

Bloomberg

Data infrastructure. $12B revenue. 325K terminals. The standard because everyone uses it.

Art

MIRAS

Computational identity + compliance layer. Visual DNA fingerprint. Blockchain-anchored. Open verification.

The data flywheel

Every registration makes the platform more valuable.
For everyone.

This is not a linear SaaS. It is a network-effect business. The value of MIRAS grows with the square of the artworks in the database — because every new record makes every existing record more searchable, more verifiable, and more useful.

MIRAS
Database
Artworks registered
Database grows
More value per search
More users adopt
↘
↙
↖
↗
  1. 1
    Artworks are registered — collectors, galleries, and dealers register artworks via Visual DNA. Each registration creates a 384-dimension computational fingerprint, anchored on Polygon blockchain.
  2. 2
    The database grows — museums contribute collection data for free (data-trade model). Public databases (Wikidata, ULAN, RKD, Europeana) are aggregated. The knowledge graph expands.
  3. 3
    Each search becomes more valuable — a larger database means more matches, richer provenance chains, better risk assessment. The AI search gets smarter. Verification gets more reliable.
  4. 4
    More users adopt the platform — insurance companies require MIRAS verification. Banks accept MIRAS risk scores. Dealers need MIRAS for AML compliance. The standard becomes self-reinforcing.

The business model

Open verification. Paid registration.
This is how infrastructure scales.

MIRAS separates the READ layer (free, public) from the WRITE layer (paid, authenticated). Verification is open because widespread adoption creates the network effect. Registration is paid because creating the identity record requires computational work and blockchain anchoring.

Open — Free

The Read Layer

  • AI-compiled search across public art databases
  • Artwork verification (re-photograph and compare)
  • Blockchain proof lookup (verify MIRAS ID on-chain)
  • Public provenance data access

Why free? Because every verification reinforces the standard. The more people verify, the more essential registration becomes.

Paid — Subscription

The Write Layer

  • Visual DNA registration (384-dim DINOv2 fingerprint)
  • Blockchain anchoring (SHA-256 hash on Polygon PoS)
  • MIRAS Passport generation (digital certificate)
  • AML scoring (ORS + ARS, 14 variables, patented)
  • Compliance reports, ALR checks, STR generation
  • Portfolio management, monitoring, API access

Why paid? Because creating identity records requires DINOv2 inference, blockchain gas, storage, and ongoing monitoring. This is the value-creation layer.

Tier strategy

Seven tiers. Three roles in the flywheel.

Each subscription tier serves a different function in the data flywheel. Some tiers generate data (growing the database). Some monetize data (generating revenue). Some do both.

Tier Price Who Flywheel role What they contribute
Search Free Everyone Adoption Top of funnel. Drives awareness. Every search trains the platform on demand patterns.
Passport €39/artwork One-time buyers Data + Revenue Each passport = 1 new artwork in the database + €39 revenue. Low friction entry point.
Collector €29/mo Individual collectors Data + Revenue 10 registrations/year. Portfolio data enriches the graph. Recurring revenue.
Professional €99/mo Advisors, small galleries Revenue Unlimited registrations. AML screening. Higher ARPU, professional use cases.
Dealer €399/mo Auction houses, large dealers Revenue Full ORS+ARS AML, ALR, EDD, STR. Highest ARPU. Compliance-driven, low churn.
Enterprise €15K–60K/yr Insurance, banks, platforms Revenue API integration. MIRAS becomes embedded in their workflow. Lock-in.
Museum Free Museums, institutions Data Bulk collection data in exchange for verification tools. 10,000+ artworks per institution. The data-trade engine.

Defensibility

Five layers of competitive moat.

Infrastructure plays are hard to replicate because the moat compounds over time. Each new artwork registered deepens every layer simultaneously.

1

Data Network Effect

Every artwork registered makes the database more valuable for every user. A competitor starting from zero faces a cold-start problem that grows worse over time.

2

Patent Protection

2 registered patents, 20 claims. Visual DNA methodology (UIBM #102026000009442) and AML compliance automation (UIBM #102026000010327). 20-year protection window.

3

Regulatory Lock-in

Once a dealer or bank uses MIRAS for EU AMLR compliance, switching means rebuilding the entire audit trail. Compliance history does not transfer.

4

Blockchain Immutability

Every Visual DNA hash is permanently anchored on Polygon. This creates a public, verifiable record that cannot be moved to a competitor. The data is on-chain, the identity is permanent.

5

Museum Data-Trade

Museums contribute collection data for free in exchange for verification tools. Millions of artwork records that no competitor can access. This is the data moat that makes Search valuable.

∞

Compounding Effect

Each layer reinforces the others. More data → better search → more adoption → more registrations → more data. The flywheel accelerates with every transaction.

Intellectual property

Protected by patents, validated by peer review.

Patent #1 — Visual DNA

Art Verification via Visual Fingerprinting & Blockchain

DINOv2 ViT-S/14 generates a 384-dimension computational fingerprint from 2 photos (front + back). Fingerprint is L2-normalized, hashed (SHA-256), and anchored on Polygon PoS. Anyone can re-photograph and verify in under 2 seconds.

UIBM #102026000009442 · 10 Claims · Regularization filed 22/06/2026

Patent #2 — AML Compliance

Automated Risk Assessment for Art Market Transactions

Dual-engine AML scoring: ORS (Owner Risk Score, 6 variables) + ARS (Asset & Context Risk Score, 8 variables). Combined score = MAX(ORS, ARS). 4-tier classification: GREEN (0–15), YELLOW (16–35), ORANGE (36–55), RED (56–100).

UIBM #102026000010327 · 10 Claims · Regularization filed 22/06/2026

Academic validation: SSRN Paper #6584859 · Total: 20 claims across 2 patents · 20-year protection window from filing date

Build the identity standard for the art market.

This is not another SaaS. This is infrastructure. The opportunity is to build the system that every gallery, insurer, bank, and museum will rely on — and to build it first.

Request Early Access →