Building NeuroTradeX — Architecting an AI-Driven Trading System

NeuroTradeX is an advanced, modular trading platform designed for real-time, explainable AI in financial and crypto markets. It was built with a focus on transparency, security, extensibility, and performance — combining technologies like LSTM, Transformers, SHAP, WebSocket streaming, Dockerized services, and real exchange integrations.

⚙️ System Overview

At its core, NeuroTradeX is composed of 5 decoupled services:

1. data-core: feature pipelines, OHLCV ingestion, sentiment processing via LLMs
2. model-engine: deep learning prediction using LSTM, Transformer, and SHAP explanations
3. exec-core: real or simulated execution (Binance, paper trading), with risk controls
4. dashboard-ui: live Next.js UI to monitor model confidence, signal history, logs
5. alert-system: async, multi-channel alerts via Telegram, Discord, and Webhooks

The entire system is open-source and hosted at: 🔗 gitlab.com/neurotradex

🧱 Modular Architecture

Each module runs independently, communicates through structured JSON files or WebSocket streams, and is deployable via Docker Compose or Kubernetes.

📊 Data Pipeline Highlights

The data-core module is responsible for ingesting:

• OHLCV data from Binance/Bybit

• Sentiment/news data from CoinDesk, Yahoo, Twitter

• Feature engineering with ta-lib, including:

  • RSI, MACD, Bollinger Bands, OBV, ATR, and custom signals

• Embedding of textual data using sentence-transformers and LLMs

def compute_macd(df):
    macd = df['close'].ewm(span=12).mean() - df['close'].ewm(span=26).mean()
    signal = macd.ewm(span=9).mean()
    return macd, signal

🧠 AI Modeling: model-engine

This module provides both supervised learning and interpretable AI via:

• LSTM and Autoformer for time series prediction
• SHAP values and LIME for feature attribution
• Support for both:
  • static offline training
  • or online/streaming signal classification

explainer = shap.Explainer(model, sample_data)
shap_values = explainer(data_point)
shap.plots.waterfall(shap_values[0])

All models export signal objects in a common schema for consumption by the exec-core.

🛡️ Execution Engine: exec-core

Built with security and precision in mind, this module handles:

• Signal validation
• Risk checking (drawdown cap, dynamic sizing)
• Execution via real APIs (Binance) or paper simulation
• CLI and Telegram fallback approvals

if risk_manager.validate(signal):
    executor.place_order(signal.asset, signal.side, size)
else:
    fallback.notify(signal, reason="risk_threshold_breached")

Trade logs and decisions are timestamped and persisted locally.

📺 UI: dashboard-ui

Written in Next.js + Tailwind, the dashboard:

• Streams signal flow and execution feedback
• Shows confidence metrics and model explanations
• Uses TradingView for OHLCV chart overlays
• Offers real-time logs and theming (dark/light)

<LightweightChart
  data={ohlcv}
  signals={executedSignals}
  overlays={shapAttributions}
/>

🔔 Real-Time Notifications: alert-system

This service asynchronously pushes alerts to:

• Telegram channels
• Discord webhooks
• Any custom webhook endpoint

It uses jinja2 templating, retries, exponential backoff, and tokenized configs via .env.

def send(signal: TradeSignal):
    message = render_template("signal_message.md", signal)
    bot.send_message(chat_id=chat, text=message, parse_mode="Markdown")

Logs are grouped by date in /logs/alerts/YYYY-MM-DD.log.

🧪 Testing & CI/CD

Each module is:

• Fully unit-tested via pytest or jest
• Integrated into .gitlab-ci.yml for linting, security checks, and test execution
• Built for containerized environments with Docker and optional support for Kubernetes Helm charts

🌍 Final Thoughts

Developing NeuroTradeX was an exercise in bringing together:

• Machine learning
• Financial engineering
• Distributed systems
• Explainability
• Human-in-the-loop design

Its open-source nature allows contributors, researchers, and traders to explore, extend, and build upon a transparent architecture built for reliability and real-world applicability.

Whether you're working in DeFi, TradFi, or AI-driven quantitative systems, NeuroTradeX provides a foundation ready to grow.

Repository: gitlab.com/neurotradex