FranceTechnology and telecom2026 – ongoingMirakl
Catalog Transformer for supplier catalog onboarding
Mirakl cut supplier catalog onboarding from 28 days to under 24 hours with generative AI
Problem
Mirakl runs an e-commerce platform for more than 100,000 merchants. Onboarding a supplier catalog took 28 days.
Approach
Mirakl built Catalog Transformer, a multi-stage generative AI pipeline on Databricks, orchestrated with Lakeflow Jobs. It serves models including GPT, Llama, Mistral and CLIP to categorize, extract, enrich and rewrite vendor product data. The company later built Mirakl Nexus, an agentic commerce layer using Agent Bricks.
Results
Takeaway
Mirakl applied the AI pipeline to the slowest step in getting suppliers live.
Categories
- Industry
- Technology and telecom
- Business function
- Production and quality
- Innovation type
- Product and service innovation
- AI technology
- Generative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes Mirakl's Catalog Transformer, a generative AI pipeline that onboards supplier catalogs. It reports onboarding time down from 28 days to under 24 hours, about 50% fewer categorization errors and 30x faster iteration for data and AI teams. It also mentions a later agentic layer, Mirakl Nexus.
Full details are in the original sources.
