Technology stack

Statistical models

Z-score, Ornstein–Uhlenbeck, cointegration methods for mean-reversion signals.

ML ensemble

Gradient boosting and regression models for signal quality filtering and false-positive reduction.

Sentiment layer

LLM-powered news and macro sentiment scoring fed as confidence modifiers into the signal ensemble.

Exposure veto

Hard caps on gross exposure per pair, sector, and total portfolio. Veto triggers halt the pipeline.

Drawdown guard

Rolling drawdown monitoring with configurable soft-warning and hard-stop thresholds.

Position sizing

Volatility-adjusted Kelly-fraction sizing with operator-configurable multipliers.

Market data layer

Normalized OHLCV ingestion from multiple providers via pluggable adapters.

News and macro feeds

Real-time and historical news feeds, economic calendar, and macro indicator streams.

Time series store

Optimized time series storage for backtesting and live signal validation.

Broker adapters

Modular adapters for broker and exchange connectivity — pluggable by environment.

Pre-trade compliance

Symbol, account, and margin checks before order submission.

Audit log

Immutable per-order audit trail for review, debugging, and regulatory compliance.

Signal dashboard

Real-time view of active signals, ensemble scores, and risk chain status.

Backtest reporting

Walk-forward performance metrics, Sharpe, Calmar, and drawdown reports.

Operator controls

Live override switches for signal gates, risk parameters, and execution toggles.