Quantitative Development (Quant Engineering)
Quantitative Development is the systems and software engineering backbone of quantitative finance. Quant developers design, build, and maintain the ultra-low-latency execution engines, order routing networks, tick-level time-series databases, and event-driven backtesting infrastructure that power billions of dollars in daily algorithmic volume.
This pillar is organised into nine topic folders (folder-per-topic), each a self-contained hub index.md plus six sub-pages walking from intuition to working formulas and code. Follow them in the order below - they climb from the research/prototyping layer, through the compiled hot path and the operating system, to the data stores, the backtest engine, the wire, and finally the live production system.
Core Quant Development Topics
- Python Quant Stack: The research-and-prototyping layer that front-ends the compiled core - NumPy vectorization, pandas pitfalls, Numba/JIT acceleration, and where the traps lie.
- High-Performance C++ for Trading: Zero-allocation paradigms, CPU cache locality (L1/L2/L3), cacheline false sharing, zero-cost abstraction, and SIMD vectorization on the tick-to-trade hot path.
- Concurrency & Lockless Programming: Why locks are slow, lock-free structures (SPSC ring buffers, the LMAX Disruptor), the C++ memory model, memory fences, and atomic synchronization.
- Low-Latency Linux & Networking: Turning a stock Linux box into a deterministic execution platform - kernel tuning, kernel-bypass NICs (Solarflare/DPDK), market-data networking, and tail-latency measurement.
- Tick-Level Databases & Time-Series: Column-oriented architectures, storage formats and compression, kdb+/q vs DuckDB/ClickHouse, and point-in-time as-of joins.
- Data Infrastructure & Reproducibility: Data pipelines, versioning and lineage, checksum arithmetic, and the reproducibility triangle that makes a backtest rebuildable bit-for-bit.
- Event-Driven Backtesting Engines: Vectorized vs event loops, order state machines, realistic fill modeling, and deterministic historical replay.
- FIX Protocol & Exchange Connectivity: Tag-value FIX, session management and sequence recovery, order lifecycles, and binary ITCH/OUCH venue protocols.
- Production Trading Systems: Lifecycle and deployment, monitoring and alerting, wire-speed risk guards and kill switches, reconciliation, and surviving contact with reality.
Reading Path (Zero to Production Trading System)
Before this pillar (foundations): read Numerical Methods first - see the Math Foundations hub for the full consumption order.
A guided route through the nine folders, in five stages.
- Start (from nothing - the prototyping layer): 1 · Python Quant Stack. Learn the research stack that expresses a signal in an afternoon - and the vectorization/pandas traps that quietly corrupt it. Then 2 · High-Performance C++ for Trading for what the compiled core is doing when the researcher’s loop gets too slow.
- Concurrency & the OS: 3 · Concurrency & Lockless Programming → 4 · Low-Latency Linux & Networking. How threads share data without locks, then how the kernel and NIC are tuned to deliver packets with a small tail.
- The data layer: 5 · Tick-Level Databases & Time-Series → 6 · Data Infrastructure & Reproducibility. Where billions of ticks live and are queried without look-ahead, and the pipeline/versioning discipline that makes the result reproducible by anyone.
- Backtesting engines: 7 · Event-Driven Backtesting Engines. Replay time honestly - signals, orders, and fills on one priority queue - before trusting any number.
- The wire & the live system: 8 · FIX Protocol & Exchange Connectivity → 9 · Production Trading Systems. How orders actually travel to the venue, and what stops the loss when the code misbehaves in production.