Skip to content

overview.md

Latest commit

 

History

History
97 lines (71 loc) · 5.32 KB

overview.md

File metadata and controls

97 lines (71 loc) · 5.32 KB
Raw

QuantumWealth Platform Overview

What Is QuantumWealth

QuantumWealth is a production-grade, AI-powered wealth management and robo-advisory platform designed to deliver institutional-grade portfolio analytics to individual investors. It combines modern portfolio theory, machine learning, and real-time market data in a fully containerized Django application.

Platform Goals

  • Provide scientifically grounded portfolio construction using mean-variance optimization, Black-Litterman, Risk Parity, and Hierarchical Risk Parity.
  • Deliver comprehensive risk analytics including Value at Risk, Conditional VaR, Monte Carlo simulation, and historical stress testing.
  • Automate tax-loss harvesting with wash-sale compliance and after-tax return optimization.
  • Generate actionable rebalancing recommendations with tax-aware trade sequencing.
  • Surface multi-signal market sentiment (news, momentum, volume, RSI).
  • Support historical strategy backtesting with transaction cost modeling.
  • Detect anomalous portfolio behavior using Isolation Forest and statistical methods.

High-Level Architecture

QuantumWealth/
|-- code/
|   |-- backend/          Django REST API, authentication, business logic, database
|   |-- ai_models/        AI and ML modules (importable Python package)
|       |-- portfolio_optimizer/    Four optimization strategies
|       |-- risk_engine/            Risk metrics and stress testing
|       |-- robo_advisor/           Goal planning and rebalancing
|       |-- market_predictor/       Price forecasting and regime detection
|       |-- tax_optimizer/          Tax-loss scheduling and after-tax modeling
|       |-- sentiment_analyzer/     Multi-signal sentiment scoring
|       |-- factor_models/          Fama-French exposure and attribution
|       |-- backtester/             Historical strategy simulation
|       |-- anomaly_detector/       Isolation Forest anomaly detection
|-- docs/                 Full project documentation
|-- infrastructure/       Nginx, PostgreSQL initialization scripts
|-- scripts/              Utility and deployment scripts

Technology Stack

Layer Technology Purpose
Web Framework Django 5 and Django REST Framework REST API, ORM, admin panel
Authentication SimpleJWT with token blacklisting Stateless auth with refresh rotation
Database PostgreSQL 16 with psycopg3 Primary data store
Cache Redis 7 via django-redis Quote caching and session data
Task Queue Celery 5 with django-celery-beat Background and periodic tasks
Portfolio Math NumPy, SciPy, cvxpy Optimization and statistics
Market Data yfinance Historical and live price data
ML Models scikit-learn Isolation Forest anomaly detection
API Docs drf-spectacular OpenAPI 3.1 schema, Swagger, ReDoc
Reverse Proxy Nginx Rate limiting, static files, SSL
Containers Docker and Docker Compose Full-stack orchestration

Key Design Decisions

Monolithic AI Integration

The AI models are native Python modules imported directly by Django, rather than a separate microservice. This removes network latency, simplifies deployment, reduces operational overhead, and makes debugging straightforward with a single process tree.

Synchronous Computation with Task Offloading

Heavy computations (Monte Carlo, backtests) are designed to run synchronously for small portfolios and can be offloaded to Celery workers for production workloads by wrapping the service call in a Celery task.

Tax-Aware Transaction Recording

Every sell transaction records cost basis, realized gain/loss, holding period in days, and long/short-term classification at execution time, enabling accurate reporting without reconstructing history later.

Redis Caching for Market Data

Live quotes are cached for 60 seconds, historical OHLCV data for 1 hour, and search results for 1 hour. This prevents rate-limit issues from yfinance and keeps the UI responsive.

User Roles and Access Control

Role Description Access Level
Anonymous Unauthenticated visitor Register and login only
Authenticated User Verified account holder Full platform access to own data
Staff Internal team member Django admin read access
Superuser Platform administrator Full Django admin access

Compliance Considerations

  • Wash-sale detection flags buy and sell of the same security within 30 days on either side of a loss sale.
  • Tax-loss harvesting recommendations include IRS-compliant substitute securities.
  • All financial computations are informational only. The platform does not constitute registered investment advice.
  • User passwords are hashed with bcrypt. JWT tokens use HS256 signing with configurable expiry.
  • Refresh tokens are blacklisted on logout to prevent reuse.