AWS data pipeline
Shopify taxonomy & category mapper
A backend pipeline that classifies any store's catalog against Shopify's Standard Product
Taxonomy — rules first, AI where it helps, with human review and write-back.
Run once across a live catalog of over 500,000 products. New products are now classified automatically as part of the daily import.
What it does
Assigning every product to the right Shopify taxonomy category is tedious and error-prone by hand, and risky to fully automate. This pipeline does the heavy lifting while keeping a human in the loop for anything uncertain.
- Classifies products in two stages: fast, deterministic rules first, then Claude AI to re-classify only what the rules flag.
- Splits results into high-confidence (ready to write) and low-confidence (flagged for human review) — no silent writes.
- Writes approved classifications back to Shopify.
- Store-agnostic: per-client configs (store domain, custom rules, CSV input), so it runs for any catalog.
- Every run produces an auditable report — counts, review splits, and an AI cost estimate.
How it's built
- A Python pipeline built around a reusable engine package — the CLI is a thin wrapper, designed to drop into a service later.
- Two-stage classifier (deterministic rules + Claude AI) with dry-run and stage-by-stage execution modes.
- Per-client configuration and rules; Shopify Admin API for write-back.
- Built for scale and trust — catalogs from a hundred to 250k+ products, explicit review states, and validation before any write.