Portfolio/Blog

Saikospeed

Performance parts + ecommerce, built with an engineering mindset: fast feedback loops, measurable results, and a clean buying experience.
Visit Saikospeed

The industry data problem

Aftermarket inventory data is messy in a way that feels structural, not accidental. Stock counts are often wrong or stale, catalogs change without warning, and every supplier publishes availability in a different format with different assumptions. There’s no universal “sync inventory to your store” API, so retailers either accept the chaos or invest the engineering effort to normalize it.

Most retailers don’t bother because the payoff is invisible when it works. But when it fails, it’s loud: oversells, cancellations, refunds, and support tickets. Inventory accuracy is not a back-office detail. It’s the foundation of customer trust.

Some vendors make it harder than it needs to be. One partner “tracked” inventory with traffic-light colors. Green meant “plenty.” Yellow meant “low, call to confirm.” Red meant “out.” No counts, no timestamps, and manually updated once per day. If only there was a thing called a digital ledger that tracks inventory in and inventory out, so you wouldn’t need a person tediously counting boxes every day like it’s 1998 🤔 Running inventory like a traffic light doesn’t scale, and they refused to improve it, so we stopped working with them.

The solution

To prevent oversells and keep the storefront trustworthy, I built an ingestion pipeline that pulls stock from multiple vendor sources, normalizes it, and updates a single source of truth database that Shopify can reliably sync from.

Vendors feed DynamoDB through separate custom Lambda sync streams, depending on what’s available:

A stock-update Lambda periodically reads DynamoDB, computes deltas, and pushes only the necessary updates into Shopify.

Architecture

UML diagram of the Saikospeed stack:

Saikospeed full-stack UML diagram

Components

Highlights

Visit Saikospeed