A deterministic multi-agent financial analysis platform that evaluates bonds, equities, and derivatives using institutional-grade valuation models and peer comparison methodologies.

Built for Hack@Brown 2026.

Traditional AI-powered financial tools achieve approximately 60% accuracy due to hallucination in mathematical calculations. Over or Under achieves 100% accuracy (26/26 tested against expert consensus) by separating deterministic financial computation from natural language generation.

Architecture:

• Financial calculations: Pure Python using established formulas
• Peer analysis: Statistical comparison against sector benchmarks
• Natural language: Claude API for human-readable explanations only

Intrinsic Value Models:

• Discounted Cash Flow (DCF)
• Graham Number (Benjamin Graham formula)
• Gordon Growth Model (Dividend discount)

Market-Based Metrics:

• Shiller CAPE Ratio
• Buffett Indicator
• Fed Model
• Rule of 20
• Tobin's Q
• PEG Ratio

Peer Comparison:

• Relative P/E positioning
• P/B, P/S, EV/EBITDA analysis
• ROE comparison

Credit Analysis:

• Credit spread calculation
• Rating consistency verification
• Yield curve positioning

Peer Comparison:

• Yield spread vs same-rated peers
• Duration-adjusted comparison

Volatility Analysis:

• Implied volatility vs historical
• Greeks comparison

git clone https://github.com/7Krisha/Over-or-Under.git
cd Over-or-Under

python3 -m venv venv
source venv/bin/activate

pip install -r requirements.txt

export ANTHROPIC_API_KEY="your_key_here"

streamlit run app.py

Over-or-Under/
├── app.py
├── agents/
│   ├── bond007.py
│   ├── stonker.py
│   ├── call_me_maybe.py
│   └── insight_generator.py
├── data/
│   ├── equities.csv
│   ├── bonds.csv
│   ├── derivatives.csv
│   └── industry_benchmarks.json
├── requirements.txt
└── README.md

• Analysis latency: Under 2 seconds
• Core calculations: Under 100ms
• 10,000x faster than traditional analyst reports

• Bond007: Fixed income specialist
• Stonker: Equity valuation expert
• CallMeMaybe: Options analysis agent

MIT License - Copyright (c) 2025 Krisha Fulgagar

GitHub: github.com/7Krisha Project: github.com/7Krisha/Over-or-Under

Built for Hack@Brown 2026