System design
Engineering notes and architecture summaries for research-to-production trading systems.
A full-lifecycle SPX 0DTE momentum system spanning backtest to live trading. Python/Rust (PyO3) hybrid with an incremental LiveState that matches the backtest engine, IBKR execution, three-tier partial take-profit, time-based exits, max-trades and consecutive-stop guards, a SQLite trade journal, and paper/live modes with safety checks.
An event-driven gamma-squeeze detector for the final fifteen minutes of the session. Rust market-data ingestion (Databento ES MBP-1 and OPRA, Polygon SPX), a signal engine for price acceleration, order-flow imbalance, sweep detection, and IV monitoring, IBKR execution, and GEX analytics.
An SPXW 0DTE sniper that fires on confluence between dealer Gamma/Vanna hedging pressure and order-book imbalance. Four-layer architecture: data adapters, a Rust Greeks engine, signal-confluence logic, and execution/risk, with an adapter pattern supporting IBKR, Databento, and Polygon.
A real-time Gamma Exposure monitor in Python and Rust with sub-100-microsecond Greeks, lock-free ring buffers, an IBKR TWS feed, WebSocket push, and a React dashboard. Exposes REST/WebSocket APIs and ships containerized with a graceful Python fallback.
An enterprise SPXW 0DTE terminal targeting sub-5-millisecond tick-to-pixel latency. Rust core (Tokio, Databento, lock-free buffers, shared-memory IPC), Python middleware (FastAPI, GEX/Greeks engines), and an Electron + React UI across three components.
A production SPXW 0DTE breakout system in Python 3.12 and Rust (PyO3) with a four-dimensional stop-loss (option value, adverse SPX move, time/progress, IV crush). Redis/PostgreSQL backend, dual IBKR connections, hot-reload config, structured logging, and golden/unit/integration tests under mypy --strict.
An IBKR TWS and Databento ingestion and real-time subscription stack for ES, SPX, and SPXW. Historical downloaders, futures stitching, L2 stress testing, rollover and chain discovery, plus a data catalog and audit tooling that serve as the data backbone for the research projects.
An execution-aware backtesting framework with strict no-lookahead timing — signals on closed bars only and dual minute/1-second clocks — plus 1-second tick replay, walk-forward out-of-sample aggregation, cost/borrow/slippage scenarios, and overfitting gates (Deflated Sharpe, PBO, bootstrap).