Software & Platforms|Product Engineering|Enable Innovation

MiraklVendor catalog onboarding

Manual reviews bottlenecked onboarding for 28 days. A GenAI pipeline now rewrites and enriches catalogs in under 24 hours.

Jan 24, 2026|12 days ago

Key results

Onboarding Time
<24 hours
vs 28 days
Fewer Errors
~50%

The company

SaaS platform for enterprise marketplaces, dropshipping, and retail media.

IndustrySoftware & Platforms
LocationParis, France
Employees1K-5K
Founded2011

Result highlights

  • Catalog onboarding cut from 28 days to <24 hours
  • ~50% reduction in categorization errors

The story

The leading provider of eCommerce software solutions supports over 100,000 merchants worldwide with rapidly growing volumes of operational and product data.

Supplier catalog onboarding was a manual process that took an average of 28 days per vendor to complete. Strict data governance regulations prevented the team from simply sending customer data to third-party AI vendors for faster processing.

The engineering team built a GenAI application called Catalog Transformer that uses a multi-stage pipeline orchestrated with Lakeflow Jobs to categorize and enrich data. The system runs a mix of models including GPT, Llama, and Mistral to rewrite vendor content while Unity Catalog maintains strict governance. Spark parallelizes the workload to process thousands of items concurrently across extensive catalogs.

Quotes

Explore similar

Find AI opportunities for your
business context

Understand what's working with 2,383 recent AI case studies across industries. We structure things so you can find high-impact strategies for your exact context.

Graphic placeholder

Industries covered