Our product turns research into routed trades. Explore how signals are generated and scored (Identification Engine), how orders are sent with guardrails (Execution Router), and how capital is protected (Risk & Controls).
Combine indicators, ML classifiers, and RL policies. Weight by performance, apply regime filters, and output one final score per asset and horizon.
Weights, half-life decay, and confidence per strategy.
Offline for training; online with Redis caches for live inference.
PyTorch, TensorFlow, XGBoost/LightGBM, scikit-learn.
Research/backtest/live in QuantConnect; LEAN-compatible.
MLflow tracking and model versions.
Final score includes components and rationale.
Step 1 — Generate signals — Build in QuantConnect or Python; emit signals with confidence.
Step 2 — Ingest & score — Normalize, weight, and apply regime/liquidity vetoes to produce a final score.
Step 3 — Publish — Expose a per-asset final_score with TTL/horizon to the execution layer.
Inputs: [signal_i, confidence_i, weight_i]
Formula: final_score = Σ(signal_i × weight_i × confidence_i) – vetoes(regime, liquidity, risk)
Output: per-asset score table with TTL/horizon.
{
"ts": "2025-09-03T12:00:02Z",
"strategy_id": "rsi_ema_v3",
"symbol": "SPY",
"insight": {"direction":"Up","magnitude":0.012,"confidence":0.73,"horizon":"1D"},
"context": {"vol":0.18,"regime":"risk_on","model":"xgb_2025_09_01"}
}
Spring Boot microservices connect to your brokers with idempotent orders, retries, and full audit. Our router consumes final_score and converts intent to venue-specific orders.
Venue adapters — Interactive Brokers, Deribit, Coinbase; roadmap: Zerodha, Saxo, Delta Exchange.
Idempotent orders — Deterministic clientOrderId, safe retries.
Resilience — Circuit breakers, clock sync, health checks.
APIs — REST/WebSocket; positions & PnL snapshots.
Audit — Who/why/when/model/thresholds/venue for every action.
When score > 0.70 and vol < 25.
Tiny sizing for low-confidence scenarios.
Enforce limits before orders leave your system—then trace every decision end-to-end. Configure exposure caps, leverage max, and per-strategy daily loss limits. Circuit breakers and a one-click kill switch halt all live flow instantly.
Global kill switch & venue-level circuit breakers
Per-strategy daily loss & drawdown limits
Hard caps: notional, size, leverage, open positions
Role-based approvals for going live
Full decision audit: inputs, model, thresholds, venue, timestamps
Yes—configure policies per symbol, strategy, and venue.
Yes—mirror orders without sending to the venue to validate policies.
Policies can be simulated to avoid surprises live.
We are software, not a broker or advisor. You remain responsible for regulatory compliance (AFM/SEBI/SEC/FINRA, etc.).