Hermes Engine

A four-phase quantitative trading system for turning research ideas into monitored, execution-ready strategies.

FastAPI BackendReact DashboardNautilus ValidationBroker Execution

Command Center

Research Runtime

Stable

Execution Mode

Paper trading

Research

Factor screening

Validation

Research gate

Scope

Private research

Portfolio Telemetry

Historical Data Service

Market data bridge into DuckDB and Nautilus

Research Gate

Five-layer validation before deployment

Execution Adapters

IBKR and Futu order path with reconciliation

System Architecture

Research to Execution Flow

Hermes separates fast research iteration from execution-sensitive validation, then pushes only validated strategies into monitored broker runtime.

Market Sources

External Data

Equities

Macro

News

Broker state

Phase 1

Data Platform

PostgreSQL

DuckDB

Parquet cache

Feature builders

Phase 2

Research Layer

VectorBT sweeps

Nautilus validation

LightGBM

Regime models

Phase 3

Deploy & Runtime

Research gate

Allocator

Risk scaling

Dashboard sync

Phase 4

Execution

IBKR

Futu

Order manager

Reconciliation

Fast path

VectorBT scans many candidates quickly before expensive event-driven checks.

Validation path

Nautilus models execution constraints before a strategy becomes deployable.

Runtime path

Dashboard, account state, orders, and risk checks stay synchronized through the backend.

Data Platform

Multi-source market ingestion, PostgreSQL storage, DuckDB analytics, Parquet cache, technical factors, geometry features, and sentiment signals.

Yahoo Finance / AlphaVantage / FRED / GDELT

Research Layer

Fast VectorBT sweeps narrow factor candidates before Nautilus Trader validates event-driven execution behavior.

VectorBT / Nautilus / LightGBM / Optuna

Deploy & Runtime

Research gates, account matching, allocation, GARCH-style scaling, and health monitoring before strategies reach live capital.

FastAPI / React / PostgreSQL sync

Execution

Broker adapters, order management, position reconciliation, online learning from fills, scheduling, and alerting.

IBKR / Futu / Scheduler / Discord

Research Gate

Strategies move through validation instead of jumping from a notebook to production.

Runtime Monitoring

Dashboard telemetry keeps equity, buying power, positions, orders, and health checks visible.

Backtest Parity

VectorBT handles rapid mining while Nautilus checks execution-sensitive behavior.