Skip to content
Case studies

Built and running

Real systems from our own companies, the first customers of everything we build. Names left out on purpose.

  • 1.2M

    products in a live ecommerce catalog

  • 124M

    fitment records in our engine

  • ~21M

    items in Google Shopping feeds

  • 109

    Claude Code skills running our own companies

  1. Online parts retailer

    A catalog of 1.2M+ products, normalized and listed with AI.

    Challenge
    About 100 brands, each sending product data in its own format. Listing it by hand would never keep up.
    What we built
    AI tooling that cleans and normalizes brand data into one catalog, then generates and maintains Google Shopping feeds.
    Result
    1.2M+ products live, and Google Shopping feeds of about 21M items at about 99 percent approval.
  2. Fitment engine

    124M+ fitment records behind every compatibility check.

    Challenge
    Shoppers need to know a part fits their vehicle, and bad fitment data means returns and rejected listings.
    What we built
    A fitment engine that maps parts to the vehicles they fit, at catalog scale.
    Result
    124M+ records powering compatibility and catalog cleanup.
  3. Parts manufacturer

    In house EDI, from purchase order to invoice.

    Challenge
    A manufacturer selling through a retailer's drop ship program had to trade X12 EDI documents, on the retailer's rules.
    What we built
    An X12 engine over AS2: purchase orders in, acknowledgments both ways, ship notices and invoices out, with price checks, a short ship guard, a dashboard and alerts.
    Result
    Live in production since September 2026. New purchase orders are picked up every two minutes.

What we're seeing

  • 01

    Catalog data is the bottleneck, not the storefront.

    Most selling problems we see trace back to product and fitment data, not the website.

  • 02

    Most companies need an AI layer on existing systems, not a rebuild.

    Agents and automations that plug into the tools you already run pay off faster than starting over.

  • 03

    Agents work when a human approves the risky step.

    Let the agent do the research and the busywork. Keep a person on the decisions that cost money.

Start a projectIndustries we build for