Muhammed Mustafayev — independent software engineer
Custom platforms for real businesses
Work across Türkiye and Europe
From product architecture to production
AI integrated into practical systems
Next.js · TypeScript · Python · Supabase
Muhammed Mustafayev — independent software engineer
Custom platforms for real businesses
Work across Türkiye and Europe
From product architecture to production
AI integrated into practical systems
Next.js · TypeScript · Python · Supabase
All work
Case Study

QT Trading OS

Institutional crypto trading engine written in Rust.

RustaxumPythonONNXNext.js

Most trading bots are a script and a prayer. QT was built the way exchanges build infrastructure: Rust at the core, a promotion pipeline in front of it, and a dead-man switch sitting above everything.

01

The brief

Prop desks and family offices do not buy signals, they buy accountability. Every prediction has to arrive with a number attached: probability, confidence, expected return, expected drawdown. And no strategy can reach live capital because someone liked the backtest chart.

02

The build

A Rust workspace of sixteen crates carries the engine, execution, risk and backtesting behind an axum API. Model training lives in Python — PyTorch, XGBoost, LightGBM, CatBoost — and exports to ONNX so inference runs inside the Rust hot path instead of over a network hop. Position sizing is fractional Kelly with volatility targeting. Validation is walk-forward plus Monte Carlo, with an online drift detector (PSI) watching for the day the market stops resembling the training set. A Next.js dashboard sits on top; Docker, Kubernetes, Prometheus and Grafana keep it observable.

03

The discipline

Every strategy walks the same corridor: research → backtest → paper → shadow → live. No shortcuts, no exceptions. A dead-man heartbeat and a kill switch sit above the whole engine with authority to stop it. The system is multi-tenant behind JWT and currently runs in private beta.

The scoreboard

  • Every signal carries probability, confidence and expected drawdown
  • research → backtest → paper → shadow → live: no stage is skippable
  • Dead-man heartbeat and kill switch above the entire engine

Want a story like this on your side?

Two lines about your problem is enough to start.