Back to Projects
ActiveHighlight

Crypto Trading Bot

Full-stack platform for automated crypto trading with self-learning AI system

Period
January 2023Present
Industry
FinTechCrypto-Trading
Roles
FullstackBackendFrontendArchitektDevOpsProduct Owner
Status
Active

Overview

Full-stack web application for cryptocurrency trading via the Binance API. Manual and fully automated trading with 8 nested rule types, portfolio management with one-click rebalancing, backtesting, global market monitoring, and a self-learning AI trading system with genetic evolution. In production for multiple years, daily use.

DDD architecture with 15 domains and generic 4-layer inheritance hierarchy. Self-learning AI trading: 3-stage detection (rule-based pre-filtering, statistical anomaly detection, LightGBM classification) analyzes ~400 symbols every 5 seconds. 96 self-learning patterns with Bayesian-EMA hybrid scoring and genetic evolution. Real-time price charts with custom D3 diagrams. Complete trading simulation with real market prices.

Tech Stack

Languages & Technologies
JavaTypeScriptHTMLSCSS
Frameworks
Spring BootAngularAngular MaterialFlywayD3Plotly.jsMVELTA4JLightGBM4jSMILEFirebase
Databases
PostgreSQLTimescaleDBRedis
Infrastructure
Docker-ComposeGitLab CI/CDKeyCloaknginxApacheLet's EncryptLinux ServerWebSocketBinance APIMaven
Concepts & Patterns
Domain Driven DesignEvent-Driven ArchitectureRESTVirtualThreadsMachine LearningGenetische AlgorithmenBayesian StatisticsCQRSMulti-TenancyPWACircuit BreakerBacktestingOAuth2

Architecture Highlights

  • Domain-Driven Design with 15 domains and generic 4-layer inheritance hierarchy
  • Event-Driven Architecture with loosely coupled services via ApplicationEventPublisher
  • Pluggable trading portal abstraction (Binance/Simulator interchangeable)
  • Bayesian-EMA hybrid scoring with Wilson Score Lower Bound for cold-start problem
  • Redis as AI state store with periodic PostgreSQL sync as fallback
  • Fully automated GitLab CI/CD pipeline with Docker deployment

Key Features

  • Self-learning AI trading system with 3-stage detection and genetic pattern evolution
  • 8 nested trading rule types with arbitrary combinability and MVEL script engine
  • Portfolio management with one-click target allocation rebalancing
  • Real-time analysis of ~400 symbols every 5 seconds
  • Global market monitoring with automatic response to rallies, crashes, and coin listings
  • Complete trading simulation with real market prices and backtesting
  • Progressive web app with push notifications and real-time updates via WebSocket