waylandzhang/ai-quant-book

Implements a modular multi-agent architecture where specialized agents handle distinct responsibilities—signal generation, risk control, and execution—rather than relying on single-model strategies. Covers production-grade concerns including backtesting pitfalls (lookahead bias, overfitting), regime detection with LLM integration, execution slippage, and operational monitoring. The curriculum spans 22 lessons across data pipelines, statistical foundations, supervised learning, and agent coordination for real-world trading deployment.

264 stars.