01 · The challenge
A proprietary trading client needed to capture funding-rate spreads between crypto exchanges - a market-neutral strategy that is impossible to run by hand, because opportunities open and close inside sub-second windows.
02 · What I built
A full-stack platform delivered to a proprietary trading client - continuously scans four crypto exchanges, ranks live arbitrage opportunities, and executes hedged positions on both sides automatically. Includes a real-time WebSocket market-data layer that holds persistent connections to multiple exchange streams, reconstructs live order books in memory, and rebroadcasts clean data to the execution engine and dashboard.
03 · The hard part
Closing a large hedged position cleanly. I designed an original execution algorithm based on the insight that, for a hedged trade, only the spread between the two legs affects profit - not the absolute price. It scores order-book levels on both exchanges against the target spread and fills both legs in a single pass, instead of leaking value walking the book.
04 · The outcome
Trigger-to-order latency under 50ms, achieved by pre-warming all market data in memory so nothing blocks at execution time. Roughly 95–99% single-attempt fill rate on large orders, against 75–85% for a conventional approach. Four exchanges orchestrated inside one process.
05 · Stack