αQuant7Alpha
system architecture · seven planes

the production stack.

q7 aegis ai is a closed-loop system. seven production planes transform raw market data into governed, attributed, adaptive trades — and feed every result back into the r-prism calibration tier, where three ai agents decide what ships next.

production architecture7 planes · 1 feedback loop
01Data Ingestion
market · sentiment · regime
MKTMarket data feeds
OHLCV across S1–S6 instruments; multi-broker normalization.
NWSNews & sentiment
Macro / micro headlines streamed into the FinBERT pipeline.
REGRegime inputs
Λ-Vol features: level, slope, dispersion, velocity.
raw frames
02Strategy Engines
6 scripts × 16 configs · 4 alpha drivers · 17 subsystems
TRNDTrend
Directional regime persistence.
BRKBreakout
Range-expansion thrust.
MRMean Revert
Statistical reversion.
SWPSweep
Stop-liquidity capture.
17 subsystems / script
Filters, structure detectors, exhaustion gates, anchored VWAP — see methodology deck.
S1–S6 strategies
Digital assets · bonds · index futures · commodity futures · spot · FX futures.
Engine attribution
Each candidate trade carries its emitting engine & subsystem stack.
candidate trades
03AI Governance Gate
all four layers must pass
AI 1XGBoost ML
Probability of profitable resolution given features.
AI 2FinBERT
Sentiment gate against streaming news flow.
AI 3Regime Match
Active regime ↔ parameter-set compatibility classifier.
AI 4Λ-Vol
Volatility PDE — 4 features, 5 states. Final regime gate.
Reject if any layer fails. Λ-Vol states: TREND_CALM · TREND_ACTIVE · MEAN_REVERT · STRESS_HIGH · TRANSITION.
approved trade
04Dynamic Risk Profiling
three layers · order flow → leverage
LYR 1Per-trade
Position size · ELU = Margin% × Leverage × SL% / 100.
LYR 2Per-session
Daily exposure cap · loss-streak circuit breakers.
LYR 3Portfolio
Aggregate ELU · cross-strategy correlation budget.
Order flow inputs map continuously to leverage — high-quality liquidity scales up; thin books scale down.
sized order
05Execution
route · place · monitor
OMSOrder management
Broker routing, partial-fill aggregation, retry policy.
POSPosition tracking
Real-time mark-to-market and unrealized P&L.
MONLive monitor
Regime re-check inside the trade; abort on regime shift.
open position
064-Leg Adaptive Exit
P1 scalp · P2 core · P3 extended · P4 runner
P1Scalp
40–50% out at first target.
P2Core
25–30% out at second target.
P3Extended
15–20% out at third target.
P4Runner
5–10% trailed by Λ-Vol envelope.
Per-engine allocations vary — TRND favors P3/P4 weighting; MR concentrates on P1/P2.
realized trade
07Telemetry
trades · attribution · drift detection
RECTrade record
Each fill, each leg, each exit reason — written to manifest.
ATTP&L attribution
Per-engine, per-subsystem, per-regime decomposition.
DRFDrift detection
Live edge vs cycle baseline → triggers next RX cycle.
cycle inputs
08AEGIS AI Calibration · Three-Agent Governance
agent 1 → agent 2 → agent 3 · go-live gate
A1Backtest Validator
Reproduces & verifies the candidate run.
A2Calibration Analyst
Emits AutoManifest (FX-tagged change set).
A3Deployment Validator
Stress-tests the manifest under live conditions.
GATEGo-Live Gate
Three consecutive PASS verdicts to ship.
nextend-to-end trade flowthree-agent governancemethodology deck