Building marketplaces with AI — field notes from production
How we run a live multi-vendor marketplace — AI product enrichment, hybrid keyword + semantic search, a custom AI assistant with real guardrails, and operations run by AI agents. Written from what actually shipped, not what a slide said.
White-label marketplace: what it is, how it works, and when it beats building your own
A plain-language guide to white-label and managed marketplaces — what "running one" actually involves under the surface (catalog ingestion, enrichment, search, per-vendor shipping, payments, compliance), which industries it fits beyond fitness, and the questions to ask any provider.
CatalogAI product enrichment: turning messy vendor feeds into a catalog that sells
Vendor data arrives broken — missing attributes, numeric IDs where brands should be, three languages in one field. How an AI enrichment pipeline fixes it at scale, what it costs, and the guardrails that keep it honest.
SearchMarketplace search that converts: hybrid keyword + semantic search in practice
Typo-tolerant keyword search finds "creatne". Semantic search finds "something for post-workout recovery". A real marketplace needs both — here's the architecture, the multilingual gotchas, and the metrics that matter.
AI assistantBuilding a custom AI shopping assistant: guardrails, grounding, and what not to do
An AI bot that recommends products it can't sell is worse than no bot. How we ground an assistant in live catalog data, chain guardrails from input to output, type the refusals, and evaluate it with a stronger model as judge.
OperationsRunning a marketplace with AI: content, SEO and ops on autopilot
A marketplace is a thousand small recurring jobs. How AI agents write and gate content, render product demo videos, audit SEO daily and review ops weekly — and the rule that makes it safe: AI authors, deterministic systems execute, humans gate.