A production-grade quantitative trading system processing market data through machine learning pipelines orchestrated by Apache Airflow for autonomous execution across multiple venues.
Architecture • Technology • Components • Infrastructure • Contact
Disuza Quantitative is a fully autonomous algorithmic trading infrastructure designed for institutional-level reliability and performance. The system integrates data ingestion, machine learning inference, risk management, and multi-venue execution into a unified, cloud-native platform orchestrated by Apache Airflow on Google Cloud Composer.
Note: This repository serves as a portfolio showcase. All proprietary trading logic, model implementations, feature engineering, and alpha-generating code remain private.
- Pipeline Orchestration: Apache Airflow on Cloud Composer for reliable workflow management
- Multi-Asset Support: Cryptocurrency markets with multi-venue execution
- Machine Learning Pipeline: Ensemble models with hierarchical gating architecture
- Risk Management: Position monitoring, drawdown controls, and exposure limits
- Observability: Comprehensive logging, metrics, and alerting infrastructure
┌──────────────────────────────────────────────────────────────────────────────────┐
│ DISUZA QUANTITATIVE │
│ Algorithmic Trading Platform │
├──────────────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────────────┐ ┌──────────────────┐ ┌──────────────────┐ │
│ │ DATA LAYER │ │ ML LAYER │ │ EXECUTION LAYER │ │
│ │ │ │ │ │ │ │
│ │ • Market Data │───▶│ • Feature Eng │───▶│ • Risk Manager │ │
│ │ • On-Chain Data │ │ • Model Infer │ │ • Order Router │ │
│ │ • Macro Data │ │ • Signal Gen │ │ • Position Mgmt │ │
│ │ • Derivatives │ │ • Ensemble │ │ • Multi-Venue │ │
│ └──────────────────┘ └──────────────────┘ └──────────────────┘ │
│ │ │ │ │
│ ▼ ▼ ▼ │
│ ┌──────────────────────────────────────────────────────────────────┐ │
│ │ ORCHESTRATION LAYER │ │
│ │ │ │
│ │ Apache Airflow (Cloud Composer) │ DAGs │ Scheduling │ │
│ └──────────────────────────────────────────────────────────────────┘ │
│ │ │ │ │
│ ▼ ▼ ▼ │
│ ┌──────────────────────────────────────────────────────────────────┐ │
│ │ INFRASTRUCTURE LAYER │ │
│ │ │ │
│ │ Cloud SQL │ Redis │ Pub/Sub │ Cloud Run │ Cloud Build │ IAM │ │
│ └──────────────────────────────────────────────────────────────────┘ │
│ │
│ ┌──────────────────────────────────────────────────────────────────┐ │
│ │ MONITORING & OPERATIONS │ │
│ │ │ │
│ │ Dashboard │ Alerting │ Logging │ Metrics │ Audit Trail │ │
│ └──────────────────────────────────────────────────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────────────────┘
Market Data Sources Orchestration & Processing Execution
─────────────────── ────────────────────────── ──────────
┌─────────────────┐ ┌─────────────────────────┐ ┌───────────┐
│ Exchange │────REST───▶│ Apache Airflow DAGs │ │ MT5 │
│ APIs │ │ (Cloud Composer) │ │ Broker │
└─────────────────┘ └───────────┬─────────────┘ └─────▲─────┘
│ │
┌─────────────────┐ ┌───────────▼─────────────┐ ┌──────────┐ │
│ On-Chain │────REST───▶│ Feature Engineering │─▶│ Signal │──────┤
│ Providers │ │ & ML Inference │ │ Router │ │
└─────────────────┘ └───────────┬─────────────┘ └──────────┘ │
│ │
┌─────────────────┐ ┌───────────▼─────────────┐ ┌─────▼─────┐
│ Macro │────REST───▶│ Cloud Run Functions │ │ CEX/DEX │
│ Sources │ │ (Inference Engines) │ │ Venues │
└─────────────────┘ └─────────────────────────┘ └───────────┘
| Layer | Technologies |
|---|---|
| Languages | Python 3.12+, TypeScript 5.0+, SQL |
| ML/Data | LightGBM, Pandas, NumPy, Scikit-learn |
| Backend | FastAPI, asyncio, Pydantic |
| Frontend | Next.js 14, React 18, TailwindCSS |
| Databases | PostgreSQL (Cloud SQL), Redis |
| Orchestration | Apache Airflow (Cloud Composer), Pub/Sub |
| Infrastructure | Google Cloud Platform, Docker, Cloud Build |
- Orchestration: Cloud Composer (Managed Apache Airflow), Cloud Scheduler
- Compute: Cloud Run, Compute Engine (Windows Server for MT5 Execution)
- Data: Cloud SQL (PostgreSQL), Cloud Storage (GCS), Memorystore (Redis)
- Messaging: Pub/Sub for pipeline communication
- ML Platform: Vertex AI (Training & Hyperparameter Tuning)
- Security: Secret Manager, Firestore
- CI/CD: Cloud Build, Artifact Registry
- Architecture: DAG-based pipelines with clear domain boundaries
- Code Quality: Type hints, comprehensive docstrings, linting (Ruff)
- Configuration: YAML-based configuration management with environment overrides
- Deployment: Immutable infrastructure with containerized services
- Monitoring: Structured logging, custom metrics, automated alerting
Production machine learning pipeline for model training, validation, and inference
- Hierarchical gating architecture with specialized expert models
- Automated hyperparameter optimization with Vertex AI
- Rolling validation with walk-forward analysis
- Feature versioning and experiment tracking
Multi-venue order execution with intelligent routing and risk controls
- Unified interface for multiple exchange protocols (REST APIs)
- Multi-account parallel execution with SQL state management
- Real-time dashboard communication and configuration sync
- Position and exposure management across venues
- P&L tracking and drawdown monitoring
- Telegram notifications for trade alerts and system status
Batch and scheduled data processing infrastructure
- REST API connections for market data ingestion
- Technical indicator computation engine
- On-chain data integration (blockchain analytics)
- Macro regime classification
Monitoring and control interface
- Live position and P&L visualization
- Trade execution history and analytics
- System health monitoring
- Configuration management
- Region: europe-west4 (GCP) with europe-west1 redundancy
- Execution Engine: Compute Engine Windows Server (MT5)
- Services: Cloud Run containerized deployments
- Orchestration: Cloud Composer managed Airflow
- Secrets management via GCP Secret Manager
- Service accounts for authentication
- Encrypted credentials storage
disuza-quantitative/
│
├── src/ # Core trading platform
│ ├── core/ # Shared infrastructure (messaging, state)
│ ├── data_pipeline/ # Market data ingestion & processing
│ ├── features/ # Feature engineering engine
│ ├── ML/ # Model training & inference
│ ├── ml_pipelines/ # Pipeline orchestration
│ ├── functions/ # Cloud Run service deployments
│ └── utils/ # Shared utilities
│
├── trading-execution-engine/ # Order execution system
│ ├── src/
│ │ ├── core/ # Configuration & data models
│ │ ├── dispatcher/ # Signal routing
│ │ ├── execution/ # Venue-specific executors
│ │ └── workers/ # Background processors
│ └── external-watchdog/ # Health monitoring service
│
├── disuza-dashboard-backend/ # Operations API
│ ├── routers/ # API endpoints
│ └── services/ # Business logic
│
├── disuza-site/ # Web dashboard (Next.js)
│ ├── app/ # Pages & layouts
│ └── components/ # React components
│
├── config/ # Configuration files
│ ├── ml/ # Model configurations
│ ├── production/ # Production settings
│ └── backtesting/ # Backtest parameters
│
├── dags/ # Airflow DAGs (pipeline orchestration)
│ ├── scripts/ # DAG helper scripts & job submissions
│ └── pipelines/ # Pipeline definitions
│
└── docs/ # Technical documentation
- Fault Tolerance: Automatic retry with exponential backoff, graceful degradation
- Observability: Structured logging, Cloud Monitoring metrics, Telegram alerts
- Scalability: Stateless services, horizontal scaling via Cloud Run
- Reliability: DAG-based orchestration with dependency tracking
This project showcases expertise in:
- System Design: Distributed systems, pipeline architecture, microservices
- Machine Learning: Production ML pipelines, model serving, feature engineering
- Cloud Engineering: GCP services, infrastructure as code, CI/CD
- Software Engineering: Python best practices, async programming, API design
- Financial Engineering: Market microstructure, execution algorithms, risk management
- DevOps: Containerization, monitoring, logging, alerting
- Full-Stack Development: React, Next.js, FastAPI, PostgreSQL
Detailed documentation is available in the docs/ folder:
Disuza Quantitative
Building the future of systematic trading.
This repository is a portfolio showcase. All proprietary trading logic, algorithms, and alpha-generating code are maintained in private repositories.
© 2025-2026 Disuza Quantitative. All rights reserved.