Service
AWS integrations & pipelines
The serverless backends behind the storefront: catalog sync from external systems, queue-based processing, and verification after every write. Lambda, DynamoDB, SQS, and S3 — sized from a hundred products to hundreds of thousands.
What this is
Keeping a large catalog in sync with an external source of truth is not a one-time import. Products are created, updated, re-priced, and restocked continuously, and the job is to make that run as a reliable, observable pipeline rather than a script someone reruns when it breaks.
What it looks like in practice
Three shipped pipelines, all running against real catalogs:
- Bulk product create & update — syncs catalogs from external sources into Shopify, verifying data integrity after every write. Runs daily in production; first put to work on a live catalog of over 500,000 products.
- Shopify taxonomy & category mapper — classifies a catalog against Shopify's Standard Product Taxonomy, deterministic rules first and AI only where they fall short, with human review before anything is written back.
- Taxonomy import enricher — a standalone service that classifies and stamps each batch inside a daily ERP-to-Shopify pipeline, without touching the importer that consumes it.
How these are built to be trusted
- Asynchronous queue-based processing, dead-letter queues, and alarms on queue depth, errors, and DLQ activity.
- Verification after every write, so problems surface immediately rather than at the end of a run.
- Conservative failure behaviour — a partial result is never written, and anything uncertain is surfaced for a human rather than guessed.
When it's the right call
- Catalog or operational data that has to stay correct at scale.
- An integration that keeps failing, or one nobody wants to touch.
- An agency with client work that runs past its AWS depth.
Is this the shape of your problem?
Describe the workflow you'd like to improve and what it would be worth to get right. Rough is fine — I'd rather hear the situation than a spec.