Hardware + Commerce Pipeline
Selling a large sports-card collection means photographing, identifying, and pricing thousands of cards. Done by hand that is months of work. We built the machine instead.
The Product
Selling a large sports-card collection means photographing, identifying, and pricing thousands of cards. Done by hand that is months of work. We built the machine instead.
Under the Hood
The pipeline starts at hardware: custom C firmware for microcontroller-driven card turntables (Raspberry Pi Pico 2 and STM32) that present each card to the camera, speaking a CRC-framed USB protocol to the host. A Python pipeline catalogs everything into Postgres, matching cards against a resumable year-by-year scrape of a reference database with front and back images. A Next.js storefront then sells the same inventory, priced 5% under eBay comps, with Stripe Checkout and a signed webhook that flips items to sold.
Deep Dive
## What we built An end-to-end card liquidation system: hardware that turns and presents cards, a pipeline that identifies and catalogs them, and a storefront that sells them, all sharing one database. ## Tech C firmware on Raspberry Pi Pico 2 and STM32F407 with TinyUSB, Python with Playwright for reference data collection, PostgreSQL with Alembic, Next.js 15 with Prisma and Stripe Checkout, packaged Windows tooling for the scanning station. ## Results The manual bottleneck is gone: cards flow from turntable to catalog to a live storefront where a Stripe webhook retires inventory the moment it sells. ## What is next Condition grading assistance from the captured images is the obvious multiplier.
The Outcome
Custom C firmware for 2 microcontroller turntable rigs with a CRC-framed USB protocol
12.5k lines of Python cataloging pipeline
resumable year-by-year reference scrape with front and back images
Postgres catalog with migrations
storefront priced 5% under eBay with Stripe Checkout
signed webhook marks inventory sold automatically
Keep Exploring
