Back to Projects

TraderAI AI Quant Trading System

WatonGroup
N/A
Chief Architect / Overall Project Lead
8 microservices with improved strategy backtesting efficiency

Project Overview

The project aims to provide users with trading signals and investment advice through algorithms and data analysis, helping them achieve better returns in the financial markets. Work started in June, and by November the core architecture design and the main functional modules were complete.

Technical Architecture

  • Core languages: Python & Golang & Rust
  • Web frameworks: FastAPI & Gin & Axum
  • Gateways: Caddy & Nginx
  • Message queues: NATS & Kafka
  • Databases: MySQL & Redis & Qdrant
  • Supported models: Claude, GPT, DeepSeek, Qwen, Kimi
  • Microservice architecture: multiple independent service modules

Key Features

1. System Architecture Design

The project adopts a microservice architecture, splitting the functional modules into independent services, including:

  • trader_core_service: core service
  • trader_data_center: data center
  • trader_event_handler: event handler
  • trader_event_center: event distribution center
  • trader_test_center: exam center
  • trader_trading_center: trader center
  • trader_vector_gen_service: vector generation service
  • trader_vector_retrieval_service: vector retrieval service

2. Expert Data Collection

  • Collects trading experience from human experts
  • Generates methodologies from human experts' experience and dynamically builds Workflow + Agents

3. Data Ingestion and Processing

  • Ingests multiple data sources: market quotes, news
  • Uses NATS for data communication

4. AI-Driven Trading Signals

  • Generates buy/sell signals
  • Supports expert analysis, live market watching, and post-session review

5. Workflow Design

  • Defines the Topic-Workflows that drive the system
  • Designed a complete workflow mechanism

Delivery Outcomes

  • Completed closed-loop validation of 3 core strategies from backtesting to paper trading, with win rates steady above 63%.
  • Unified the monitoring metrics system - production alert false-positive rate down 40%, critical event response time under 10 minutes.
  • Co-built a decision dashboard with the investment advisory team, enabling strategy launch reviews and joint sign-off with risk control.

Team and Governance

  • Adopted a Squad + Chapter organization model covering four groups: strategy, data, middle platform, and infrastructure.
  • Defined biweekly business reviews and roadmap syncs to keep stack evolution aligned with the pace of the business.
  • Introduced cost and performance baselines, with quarterly reviews of resource utilization and budget execution.

Usage

Core Service Modules

# 核心服务
git clone git@github.com:DescartAI/trader_core_service.git

# 数据中心
git clone git@github.com:DescartAI/trader_data_center.git

# 事件处理器
git clone git@github.com:DescartAI/trader_event_handler.git

Feature Highlights

  1. Exam center: supports trading knowledge tests and competency assessment
  2. Trader center: formatted output of buy/sell signals
  3. Vector services: provides vector generation and retrieval
  4. Event-driven: message-driven architecture based on NATS

Project Outcomes

  • Completed the system's core architecture design
  • Implemented data ingestion, signal generation, account management, and other features
  • Built a complete microservice system
  • Laid the foundation for subsequent productization and commercialization