PitLot

Collectors keep a garage, scan models, build wishlists, list for sale or auction and arrange local trades. Underneath sits a separate catalog data layer — 122,035 releases and 13,929 castings across brands, built by a dependency-free Python pipeline with 295 tests.

The Problem

Die-cast collectors trade in Instagram comments and WhatsApp groups, with no shared catalog to say exactly which casting, series and variant is on the table.

The Solution

A real catalog first, then the marketplace on top: every listing, wish and trade points at a catalog record, and the rules live in the database with tests that prove them.

What I Built

• →Garage, scanning and wishlists
• →Listings, auctions and local trades
• →Payments, orders, messaging and notifications
• →Admin resources
• →Supabase schema with pgTAP behaviour and permission tests
• →Design system with 157 tokens and 43 auditable rules

Highlights

• →122,035 releases · 13,929 castings
• →Behaviour and permission tests in the database
• →43 auditable design rules

Outcome

• →Catalog data layer complete with 295 tests
• →Pre-launch — production acceptance in progress