Patent 28 / NFT Auto-Segmentation
01 / 11 US12153794B2
↑↓ navigate  ·  all patents →
Siten Sanghvi & Naga Vamsi Krishna Akkapeddi  ·  Granted Nov 26, 2024

NFT Auto-Segmentation

A two-model ML platform for NFT lifecycle management. The auto-segmentation model scores each NFT using owner profile, account context, network conditions, and ownership history — routing it to distributed ledger storage or local storage based on a tier threshold. The validation model gates event processing requests by scoring their validity. Both models self-refine in iterative feedback loops. Social media signals update scores in real time.

US12153794B2Patent
Dec 1, 2022Filed
24 monthsTime to grant
20 Claims / 3 independentScope
2 InventorsAkkapeddi, Sanghvi
SCROLL TO EXPLORE
Visual patent explainer
02 / The Problem

Not all NFTs warrant the same storage infrastructure — but existing systems treat them identically. The result: bloated ledgers and underserved assets.

Distributed ledger storage for NFTs — blockchain or similar infrastructure — is computationally and economically expensive. Storing every NFT on-ledger regardless of its value, liquidity, or usage frequency wastes resources on low-activity assets while potentially under-provisioning high-value ones. There's no automated mechanism to dynamically route NFTs to the appropriate storage tier based on real-time signals about their value and activity.

Uniform Storage for Non-Uniform AssetsHigh-value, actively traded NFTs and dormant low-value tokens stored identically — no cost or performance optimization based on activity or value signals
No Dynamic Re-TieringEven if initial routing is correct, NFT value and ownership change over time — static storage assignment can't adapt as market conditions shift
No Automated Validation GateEvent processing requests (transfers, listings, metadata updates) require validation before execution — without an ML validation layer, this is either manual or absent, creating fraud and error exposure
03 / The Invention

Two ML models, one platform: auto-segmentation routes NFTs to the right storage tier; validation gates event processing. Both refine continuously.

The platform runs two distinct machine learning models. The auto-segmentation model takes NFT information — owner profile, account data, network conditions, geo signals, ownership history — and produces a tier score. If the tier score meets or exceeds the threshold, the NFT is stored on the distributed ledger (cloud/high-value tier). If below, it's stored locally (low-activity tier).

The validation model operates on event processing requests: when a transfer, listing, or other event is requested for an NFT, the model scores the validity of that request and generates a validation rating. Requests below the validity threshold are gated — the user receives a fallback UI notification (Claim 4). Both models iterate in feedback loops: each outcome refines the next scoring cycle, improving segmentation and validation accuracy over time.

04 / Two-Model Architecture

Auto-segmentation for storage routing. Validation for event gating. Independent models, shared platform infrastructure.

The two models solve different problems in the NFT lifecycle. Auto-segmentation is a storage optimization problem: given an NFT's current signals, where should it live? Validation is a security/quality problem: given an event request, should it be processed?

Claim 7 gives the broadest definition of what an NFT can be in this context: digital signatures, digital assets, domain names, digital artwork, account information, transaction information, or digital currency. The platform is not limited to image-based NFTs — it encompasses any digital token with ownership semantics.

Two-Model Platform — US12153794B2

Auto-Segmentation Model

  • Inputs: user info, account info, network conditions, geo info, ownership history
  • Scores: generates tier score per NFT
  • Routes: ≥ threshold → distributed ledger; < threshold → local storage
  • Re-tiers: score changes trigger storage migration
  • Social: live social media signals update score (Claim 3)
  • Iterates: each outcome refines next scoring cycle

Validation Model

  • Inputs: event processing request for an NFT
  • Scores: generates validation rating for the request
  • Gates: valid requests proceed to event processing
  • Fallback: invalid requests trigger user UI notification (Claim 4)
  • Profiles: constructs and links user profiles (Claims 12–13)
  • Iterates: outcomes feed back into model refinement
05 / Dynamic Re-Tiering

Tier scores change as NFT value, ownership, and market conditions change — and storage location moves with them.

The system is designed for dynamic storage migration. Claim 10 specifies the triggers: when an NFT's tier score changes due to network conditions, NFT value change, ownership transfer, or market condition changes, the system migrates the NFT between storage tiers accordingly.

Claim 8 specifies that training data for the auto-segmentation model includes multiple historical tier scores over time for the same NFT — not just a snapshot. This enables the model to learn from NFT lifecycle trajectories: tokens that start low-tier and move up, high-value tokens that become dormant and move down, and the signal patterns that precede each transition. The historical series is the feature, not just the current state.

Re-Tiering Triggers — US12153794B2

Network Conditions

Ledger congestion, gas costs, or network degradation can make local storage temporarily preferable even for high-value NFTs. Score adjusts, storage migrates.

NFT Value Change

Market appraisal of the NFT increases or decreases. High-value NFTs move to distributed ledger for security and liquidity. Low-value tokens drop to local storage.

Ownership Transfer

NFT transfers to a new owner whose profile scores differently. New owner context (account signals, geographic data, transaction history) triggers re-evaluation.

Market Conditions

Broader market signals — collection floor prices, trading velocity, social media activity — affect the tier score calculation. Score updates propagate to storage decisions.

06 / Tier Simulator

High-tier to ledger, low-tier to local, score change triggers migration, social signal updates validation.

Select a scenario to trace the auto-segmentation or validation pipeline from input signals through storage routing decision.

NFT Tiering Scenarios
07 / Social Signals

Claim 3: social media API integration — live sentiment and trending signals update tier scores in real time without manual reprocessing.

Claim 3 extends the auto-segmentation model with social media signal integration. The platform connects to social media APIs to pull real-time signals: mention velocity, sentiment, trending status, and influencer engagement for NFTs or NFT collections. These signals feed directly into the tier score calculation.

This creates a live feedback loop between social activity and storage routing. An NFT collection that goes viral — rapid mention increase, positive sentiment spike, sudden trading volume — will see tier scores rise in real time, with high-demand tokens automatically migrating from local to distributed ledger storage before the demand peak hits. The system anticipates storage needs from social signals rather than reacting after value has already shifted.

Social Signal Integration — US12153794B2

Mention Velocity

Rate of new social mentions for the NFT or its collection. Sudden acceleration is a leading indicator of demand increase — tier score responds before market price catches up.

Sentiment Analysis

Positive/negative/neutral classification of social mentions. Sustained positive sentiment signals long-term value appreciation; negative signals potential devaluation.

Trending Status

Whether the NFT or collection is trending on connected platforms. Trending status can spike tier scores dramatically — ensuring high-availability storage for surging demand.

Live Score Updates

Social signals update tier scores continuously via API polling. No batch reprocessing required — the auto-segmentation model receives a live social signal feed as an input feature.

08 / Applications

Adaptive storage infrastructure for any digital asset platform where asset value and activity patterns change dynamically over time.

The two-model architecture — segmentation routing + validation gating — applies to any digital asset system where cost optimization, fraud prevention, and dynamic re-tiering are simultaneous requirements.

Use Cases — US12153794B2
Express
Digital Art Marketplace Platform hosts 10M+ NFTs. Auto-segmentation routes the top 2% by activity score to distributed ledger — ensuring high availability for actively traded pieces. The remaining 98% store locally — dramatically lower infrastructure cost. When a low-tier artist goes viral on social media, their tokens auto-migrate to ledger before demand spikes.
Express
Financial Asset NFTs Claim 7's broad NFT definition includes account information and transaction information as NFT-representable assets. A financial institution uses the platform to manage digital certificates, tokenized financial instruments, or account ownership tokens — same two-model architecture, same dynamic tiering logic.
Express
Fraud-Gated Event Processing Transfer request arrives for a high-value NFT. Validation model scores the request: unusual velocity, new device signature, mismatched geo profile. Validation rating below threshold. Request gated — user receives fallback UI. Legitimate owner notified of attempted unauthorized transfer. No on-chain transaction executed.
Inferred
Domain Name Portfolio Management Domain names (Claim 7 NFT type) in a large portfolio. High-value premium domains stored on distributed ledger with full provenance chain. Parked/dormant domains stored locally. Value change in secondary markets triggers re-tiering. Social signal integration detects when a domain becomes trendy and upgrades its storage tier proactively.
09 / Citations

Forward Citations

No forward citations found as of this check. US12153794B2 was granted November 2024 and has a continuation pending (US20250028443A1, filed Oct 9, 2024). Citation data is still accumulating for this recently-issued patent.

Forward citation check: Jun 22, 2026 · Static fetch; Google Patents citations are JS-rendered
Forward Citations — US12153794B2
No citations available yet Granted Nov 2024 — citation index still populating A continuation application is pending (US20250028443A1). Check Google Patents for current forward citations and continuation status.
10 / NFT Scope

Claim 7 defines the broadest NFT definition in this patent portfolio — extending well beyond digital art.

Claim 7's definition of NFT includes: digital signatures, digital assets, domain names, digital artwork, account information, transaction information, and digital currency. This is deliberate claim drafting — the patent's scope is not limited to the image-collection NFTs that dominated popular discourse in 2022. It covers any digitally unique, ownership-tracked token that the ML platform can score and route.

A continuation application (US20250028443A1, filed Oct 2024) extends the platform further — building on the foundation of this patent with additional claims coverage. The combination of parent grant and pending continuation gives the platform architecture broad and growing protection.

What Counts as an NFT — US12153794B2

Digital Artwork

The most commonly recognized NFT type. Images, video, music, generative art. Tier score driven by social signals, trading volume, collection floor price.

Domain Names

Web3 domain names (ENS, Handshake, etc.) with unique ownership. Tier score driven by length, keyword value, secondary market demand, trending topics.

Account & Transaction Info

Financial account ownership tokens and transaction records represented as NFTs. Enables the two-model platform to operate in financial infrastructure contexts.

Digital Signatures & Currency

Unique digital attestations and tokenized currency instruments. Claim 7's breadth ensures the platform architecture applies to future digital asset types not yet standardized.

11 / Timeline

Patent Lifecycle

Dec 1, 2022
Filed
Application filed — B2 patent
18 months
Jun 6, 2024
Published
Pre-grant publication US20240184859A1
6 months
Nov 26, 2024
Granted
US12153794B2 granted — 24 months from filing
~18 years
Mar 9, 2043
Expires
Adjusted expiration with Patent Term Adjustment
End / Patent 28