

The problem
To compare second-hand gold jewellery, I had ten tabs open, worked out the price per gram by hand and kept a spreadsheet I’d end up losing.
My role
Product design, scrapers, data quality checks, interface and launch, on my own.
Decisions behind the product
Treat data as product work
Collectors feed a shared dataset from inconsistent sources. Weight, gold purity and price need to be interpreted before becoming comparison criteria.
Make price per gram visible
The interface centres price per gram and filtering. The calculation offers a common reference; it does not replace reading the listing and seller conditions.
Check the comparison
Quality checks are part of the product, just like the interface. A readable ranking only makes sense if the underlying data is consistent.
The stack
- Angular SSR
- FastAPI
- Python
- MongoDB
- scrapers
- Vercel
- Railway
The outcome
Live: a price-per-gram ranking whose reliability is tested, not assumed.