I build production quant systems — probabilistic models, risk management, and the boring-but-load-bearing operational scaffolding around them. Currently working on prediction-market trading and the Claude / Cowork plugin ecosystem.

GEO — The Kalshi Weather Bot (DEMO)  ·  live preview: apeabody007.github.io/GEO-The-Kalshi-Weather-Bot-DEMO

A live algorithmic trading system for Kalshi weather prediction markets. Multi-model probabilistic forecast (ECMWF + GFS + HRRR + NBM + GraphCast + AIFS) with per-station EMOS bias correction, fractional-Kelly sizing under stacked risk caps, maker-only pricing, signed REST client, SQLite trade ledger, a server-rendered PWA dashboard, ~110 pytest cases, and reconciliation against the exchange's official settlement source. Public repo is a read-only preview; source is private.

Quant Toolkit  ·  the framework, abstracted

A Cowork plugin distilled from operating GEO. Nine skills: Kelly sizing, calibration audit, backtest harness, EMOS bias correction, maker pricing, P&L attribution, market scanning, pre-flight checklist, drawdown monitor. Venue-agnostic — works against any binary contract priced in [0, 1]. MIT.

Python 3.11. Probabilistic forecasting (EMOS / Platt scaling / ensemble blending). Market microstructure (maker-taker fees, adverse selection, order-book dynamics). Risk management (fractional Kelly, drawdown control, position caps). Backtesting with honest execution assumptions. Reliability engineering for systems that handle money (single-instance locks, circuit breakers, structured pre-flight checks). SQLite. Tailscale-private PWAs. Claude Agent SDK + Cowork plugin development.

LinkedIn for hiring inquiries or collaboration.