AI-First Visual Intelligence Platform

Multimodal computer vision that turns physical returns into priced, listed inventory at liquidation scale.

rScan converts a raw item scan into a taxonomy-compliant, competitively priced marketplace listing in a single automated pass — then distributes it across eBay, Amazon, Walmart, Shopify and Google Shopping with inventory held in sync.

  1. 01 Physical scan
  2. 02 Multimodal feature extraction
  3. 03 Catalog synthesis
  4. 04 Predictive valuation
  5. 05 Marketplace distribution
Request Platform Access Explore Our AI Architecture
  • Accelerator Backed by gener8tor / gBETA
  • Flywheel Fund Portfolio
  • Built for High-Volume Resellers & Enterprise Liquidators
768K+ Physical item scans processed
1.78M+ Secondary-market price observations
183K+ Marketplace listings synthesized
131K+ Orders fulfilled through the platform

Production telemetry from the rScan platform database, September 2026. Serving 172 tenant operations on one multi-tenant deployment.

Core Technology & Architecture

Four technology layers, one automated pass.

Every layer below runs in production today. Learned components enrich a deterministic foundation, so an inference failure degrades a record gracefully — it never blocks receiving.

01

Multimodal Computer Vision & Feature Extraction

Identification starts at the edge and finishes in the cloud. Identifier decode runs on the operator's device; server-side vision and multimodal inference recover the attributes a marketplace actually needs.

  • On-device identifier recovery — UPC, EAN, GTIN and Amazon FNSKU from the live camera, with no network round trip.
  • Image intelligence — deduplication, authenticity checks that separate genuine unit photography from stock renders, and provenance tracking for every asset.
  • Fine-grained attribute extraction — brand, model, condition, colorway, dimensions and weight resolved from imagery and unstructured source data.
  • Edge decode
  • Image integrity
  • Multimodal extraction
  • Vision-capable models
02

Generative Listing & Catalog Synthesis

A photo, a spec sheet or a retailer page becomes a normalized, schema-validated catalog record — checked against each marketplace's rules before anything is published.

  • Structured output — every generated record is validated against a schema before it reaches inventory.
  • Taxonomy compliance — automatic classification against the Google Product Taxonomy and per-channel category trees, with brands canonicalized across 3,862 manufacturers.
  • Channel policy enforcement — prohibited terms, restricted values, trademark handling and title limits applied per marketplace, pre-publication.
  • Provider-portable inference — runs across Anthropic, Google and OpenAI models behind a single interface.
  • Schema-validated output
  • Taxonomy classification
  • Brand normalization
  • Policy compliance
03

Predictive Valuation & Liquidation Intelligence

Pricing is a learned policy, not a rules table. A reinforcement-learning model trained on rScan's proprietary corpus of secondary-market observations selects prices that balance realized margin against sell-through velocity.

  • Proprietary observation corpus — 1.78M+ secondary-market price observations, growing by roughly 183,000 per month.
  • Comparable quality enforced ahead of the model — noisy matches never shape a price.
  • Cost-floor guardrail — every price is bounded by a computed landed-cost floor, so a learned action cannot price below contribution margin.
  • GPU-accelerated training — trained weights are version-controlled and promoted deliberately.
  • Reinforcement-learned pricing
  • 1.78M observation corpus
  • Cost-floor guardrails
  • CUDA training
04

Cloud & Edge Scalability

Inference runs where it belongs — decode on the device, extraction and valuation in the cloud — on a multi-tenant platform that runs continuously in production.

  • Real-time progress streaming — operators watch a scan resolve instead of waiting on a request to return.
  • Concurrency-safe marketplace mutations — with a durable, auditable event ledger behind every create, revise and end.
  • Provider quota governance — keeps interactive store work responsive under published marketplace API ceilings.
  • Multi-tenant at scale — 172 tenant operations, 88,775 products and 183,881 listings on one deployment.
  • Edge + cloud split
  • Real-time streaming
  • Auditable ledger
  • Multi-tenant
Read the architecture overview

Enterprise & Power-Seller Capabilities

Built for throughput, measured on operational ROI.

rScan is not a listing assistant bolted onto a spreadsheet. It is the system of record for the physical unit — from the moment it is scanned on a receiving dock to the moment its margin is booked.

ThroughputBulk Scanning Workflows

One capture surface spans handheld scanning in the aisle and station-based bulk intake at a receiving dock. Operators scan continuously; the platform resolves, enriches and queues distribution behind them.

  • Web, iOS and Android from a single application.
  • Shelf assignment, audit, pick routing, returns and disposal modeled end to end.
  • 768,175 physical item scans processed to date.

IntegrityOmnichannel Inventory Synchronization

A unit listed on five channels is one unit. Quantity reconciles in real time, so a sale on one channel withdraws availability everywhere else before a second buyer can transact against it.

  • Concurrent updates are serialized — never last-write-wins.
  • Every create, revise and end recorded in an auditable ledger.
  • Per-channel compliance checks applied before publication.

MarginAutomated Order & Margin Analytics

Orders from every channel land in one queue with centralized fulfillment, rate-shopped labels and delivery tracking. Cost of goods follows the unit from acquisition to sale.

  • Contribution margin as a property of the unit, not a month-end reconstruction.
  • Daily production, hourly throughput, employee, returns and shelving analytics in-platform.
  • 131,587 orders fulfilled through the platform.

Technical Specifications & System Integration

Measured in production, not in a benchmark harness.

Identifier resolution latency

Sub-second

On-device decode returns a UPC, EAN, GTIN or FNSKU from the live camera with no network round trip.

Structured attribute coverage

98%+

Measured across the live 88,775-product catalog: 100% category, 99.99% GTIN, 99.98% title, 99.8% imagery, 98.4% normalized brand.

Distribution channels

5 live

eBay, Amazon, Walmart, Shopify and Google Shopping — each a first-party API integration.

Price observation corpus

1.78M

Proprietary secondary-market comparables, accruing at roughly 183,000 per month.

Brand taxonomy

3,862

Distinct manufacturer brands canonicalized from free-text source variants.

Multi-tenant footprint

172

Independent tenant operations isolated at the data layer, on one deployment.

Orders fulfilled

131,587

Through the platform to date, with 183,881 marketplace listings published across five channels.

Capture surfaces

Web + iOS + Android

One application, one capture and inference contract across every surface.

First-party marketplace & logistics integrations

  • eBay
  • Amazon SP-API
  • Walmart Marketplace
  • Shopify
  • Google Shopping
  • Shippo
  • Veeqo
  • QuickBooks

Platform & inference stack

  • Node.js
  • PostgreSQL
  • Angular
  • TensorFlow
  • Anthropic Claude
  • Google Gemini
  • OpenAI
  • Redis
  • Socket.IO
  • CUDA

Evaluating rScan for an enterprise liquidation operation?

We will walk your team through the platform, the data assets behind the valuation model, and a throughput model built on your actual unit volume.

Schedule an Enterprise Demo Explore Our AI Architecture