Investment System

Autonomous daily trading simulation with ML-powered market regime detection and asset health scoring.

Current Status

• Pure simulation — There is no broker. Each night the pipeline decides trade intents; the next morning it simulates the fill at the real market OPEN, and that simulated fill becomes the new forward point on the champion line.
• Continuous forward line — The displayed champion line advances one trading day per trading day, automatically. Tonight's analysis settles this morning's provisional point to the real close.
• Staged challenger — The optimizer identified macro_downgrade_threshold = -0.75 (from default -0.50) as a candidate improvement. This is staged only — not promoted to active parameters.
• Active params — opt-bootstrap bundle. No parameter changes have been made since the initial deployment.

Features

• Daily Pipeline - Automated data ingestion, feature engineering, and simulated trade execution
• ML Models - GRU/Transformer for regime classification, Autoencoder/VAE for health scoring
• LLM Integration - Claude Haiku (Bedrock) risk assessment and market weather reports
• React Dashboard - Portfolio metrics, charts, evidence surfaces, and market analysis
• Evolutionary Optimization - Genetic algorithm for policy parameter tuning
• Simulated Execution - Trade intents decided overnight are filled in simulation at the next market open; no broker is involved

Architecture

Daily (10 PM ET):
  EventBridge → Lambda → [Ingest → Features → Inference → Decisions] → S3

Monthly (1st, 2 AM):
  launchd → train.py → Models uploaded to S3

Dashboard:
  S3 Static Site → React App → Reads from S3 artifacts

Quick Start

Prerequisites

• Python 3.11+
• AWS CLI configured
• API keys: OpenAI, FRED (free tiers available)

Local Development

# Create virtual environment
python -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt
pip install -r requirements-training.txt  # For ML training

# Run tests
pytest tests/ -v

Deploy to AWS

See docs/DEPLOY.md for full deployment instructions.

# Quick deploy
./infrastructure/s3_setup.sh investment-system-data us-east-1
./infrastructure/secrets_setup.sh us-east-1
./infrastructure/lambda_deploy.sh investment-system-daily-pipeline investment-system-data us-east-1
./infrastructure/eventbridge_setup.sh investment-system-daily-pipeline investment-system-data us-east-1

Run Dashboard Locally

cd frontend
npm install
npm run dev
# Visit http://localhost:5173

Project Structure

├── src/                    # Lambda function code
│   ├── handler.py          # Main Lambda entry point
│   ├── steps/              # Pipeline steps
│   ├── models/             # Baseline + trained model loaders
│   └── utils/              # S3 client, logging
├── training/               # ML training code
│   ├── models/             # GRU, Transformer, Autoencoder, VAE
│   ├── utils/              # Data loaders, metrics
│   ├── train.py            # Training orchestrator
│   └── train_*.py          # Individual model training
├── evolution/              # Evolutionary optimization
│   ├── genome.py           # PolicyGenome class
│   ├── fitness.py          # Backtest-based fitness
│   ├── genetic.py          # Genetic algorithm
│   └── promotion.py        # Template promotion
├── frontend/               # React dashboard
│   ├── src/components/     # UI components
│   └── src/types/          # TypeScript interfaces
├── automation/             # Operational scripts
│   ├── run_training.sh     # Monthly training wrapper
│   ├── check_costs.py      # AWS cost monitoring
│   └── check_pipeline.py   # Pipeline health check
├── infrastructure/         # AWS deployment scripts
├── tests/                  # Test suites
└── docs/                   # Documentation

Documentation

• Deployment Guide - AWS deployment procedures
• Training Guide - ML model training
• Frontend Guide - Dashboard development
• Operations Guide - Day-to-day operations
• Architecture Plan - Implementation plan and status

Tests

# Run all tests
pytest tests/ -v

# Run specific test files
pytest tests/test_ml_models.py -v      # ML model tests (20)
pytest tests/test_evolution.py -v       # Evolution tests (28)
pytest tests/test_baseline_models.py -v # Baseline tests

Current test count changes over time; run pytest tests/ -v for the latest totals.

Configuration

Environment Variables

Variable	Description
S3_BUCKET	S3 bucket name
AWS_REGION	AWS region
INVESTMENT_SLACK_WEBHOOK	Slack webhook for alerts
INVESTMENT_ALERT_EMAIL	Email for SES alerts

AWS Secrets

Stored in Secrets Manager:

• investment-system/openai-key
• investment-system/fred-key
• investment-system/alphavantage-key

Cost Estimate

Service	Monthly
Lambda	~$6
S3	~$1
Secrets Manager	~$1
CloudWatch	~$1
Total	~$9

License

Private project - not for distribution.