01
Data Integrity
A strategy is only as honest as the data it was allowed to see.
- Point-in-time data layer so features only use information available at decision time.
- Schema and data-quality contracts run before any research step.
- Dedicated tests that reject look-ahead leakage instead of trusting convention.
02
Validation Methodology
Backtesting is treated as a discipline, not a marketing exercise.
- Purged walk-forward cross-validation to stop overlap between training and test windows.
- Out-of-sample evaluation feeds a PASS / FAIL policy gate before anything is deployed.
- A committed fail-demo config proves the gate rejects weak strategies, not just approves good ones.
03
Cost-Aware Backtesting
Edge that disappears after costs was never edge.
- Backtests model spread, slippage, commission, execution delay, and liquidity caps.
- Rule complexity stays constrained and results are interpreted conservatively.
- Consistency and drawdown matter more than headline returns.
04
Risk & Live Controls
Risk controls are part of the system design from the start, not an afterthought.
- Position sizing, stop-loss logic, and portfolio-level allocation limits.
- Risk metrics implemented in code: Sharpe ratio, VaR, Expected Shortfall, and max drawdown.
- Shadow and paper deployment, drift detection, human approval, and a kill switch.
05
Execution & Stack
Freelance execution work since 2021, built to respond reliably in real time.
- Event-driven architecture keeps data intake, signal evaluation, and order execution modular.
- Exchange API integrations including Binance, KuCoin, and Gate.io.
- Python for research and orchestration, Go for performance-critical services, Docker and Linux for reproducible deployment.