CryptoQuant – Quantitative Trading Platform
Production-ready algorithmic trading platform for Binance USDM Futures with 6 quantitative strategies, ML signal grading (XGBoost + GRU ensemble), and 1,261 unit tests.
System Architecture
Overview
A production-ready algorithmic trading platform for Binance USDM Futures with 6 quantitative strategies, ML-powered signal enhancement, and comprehensive backtesting.
Problem Statement
Manual crypto trading is emotionally driven and inconsistent. Traders need systematic, data-driven approaches with proper risk management.
Solution & Approach
Event-driven architecture with Redis Pub/Sub. Dual exchange connectivity (ccxt.pro WebSocket + REST). ML ensemble (XGBoost + GRU) for signal grading with ONNX runtime.
Key Features
- 6 quantitative strategies: Smart Money, Momentum, Mean Reversion, Volume, Funding Rate, Breakout
- ML signal grading (A/B/C/D) with XGBoost + GRU ensemble via ONNX
- Backtesting with walk-forward validation and Monte Carlo simulation
- Risk management: Kelly Criterion sizing, circuit breaker (5% daily / 15% drawdown)
- 1,261 unit tests + 37 E2E tests
Results & Impact
All safety, reliability, security, and data integrity checks passing. 1,261 unit tests and 37 E2E tests.
Lessons Learned
Dual exchange connectivity is necessary due to real API quirks. Signal grading with ensemble ML outperforms single-strategy systems.