How I Built a Dijkstra-Based Money Routing Engine with 29K Edges

The problem nobody talks about A nurse in the Philippines sends $300 home every month. She's used the same service since 2012. She doesn't know that a two-step route through a stablecoin she's never heard of would save her $18 every month. Over twelve years, that's $2,500 — not stolen, just invisible. I wanted to build something that shows all the paths. Not a transfer service. Not a wallet. Just a map. That's Coinnect — an open-source routing engine that finds the cheapest way to move money between any two currencies using any combination of fiat, crypto, and P2P networks. Modeling global money as a graph The core insight: every exchange, remittance provider, and P2P market is just an edge in a directed graph. Nodes = currencies (USD, MXN, BTC, USDT, NGN, GBP...) Edges = conversion paths, each carrying: - exchange_rate - fee_pct - estimated_minutes - provider name Right now the graph has 29,000+ edges across 50+ currencies, pulled from 45+ live data sources every 3 minutes: 21 crypto