Building a Production Quant Trading Stack

The book uses a real system, codenamed Titan, as a running, sanitised case study: the reasoning, the trade-offs, and the things that went wrong alongside the things that worked.

What the book covers

Part 1: Foundations establishes why most quant systems fail in production before a single trade is placed, and walks through the architecture, stack choices and project layout that survive contact with live markets.

Part 2: Research covers the discipline of building a backtest you can actually trust: deflated Sharpe ratios, a failure-mode catalogue, the full metric suite, walk-forward validation, and the Sanctuary decision matrix for deciding when a strategy is ready.

Part 3: Data addresses sourcing and storage for systematic strategies, and the data quality gate that keeps bad ticks out of research.

Part 4: Research to Production bridges the gap between a working backtest and a live strategy, covering broker realities, the live-equals-research invariant, and a typed strategy-class contract that enforces consistency across both environments.

Part 5: Portfolio and Risk covers portfolio risk management, per-strategy equity tracking with FX conversion, position sizing via a modified Kelly criterion, a correlation dial for the allocator, and a layered safety system.

Part 6: Deploy and Operations walks through containerisation, paper-trading to live transition, and a live runbook for day-to-day operations and incident response.

Part 7: Reflections closes with an honest accounting of open problems and a curated reading list.