Investment Reasoning Architecture
MARY
MARY reasons the way a CIO thinks — across macro regimes, portfolio construction, fundamental analysis, and quantitative strategy. Four specialized agents. Real data. Grounded answers.
What MARY Is
MARY is an investment reasoning architecture — a system built to analyze the questions that matter most to institutional investors: What is the macro regime? Which stocks are worth owning? How should a portfolio be constructed given current conditions? Which strategies work in this environment?
Every answer is grounded in real data. MARY pulls from 21 FRED macroeconomic indicators, SEC filings, and live market prices, and runs deterministic computations — regime scoring, Black-Litterman optimization, DCF models, backtests — before any language model synthesizes the result. The reasoning is explainable. Every claim traces back to a source.
Built and operated by Parson Tang, an investment professional with 20+ years of experience across global markets and institutional investment practice. The system reflects how a practitioner actually thinks through investment problems — not how a chatbot summarizes financial news.
The Four Minds
Each agent is a specialized reasoning system. They operate independently and in combination, depending on the question.
Agent 01
Macro Intelligence
- Regime classification across 4 states — expansion, slowdown, contraction, recovery
- 21 FRED leading indicators, z-score normalized with hysteresis buffers
- Fed NLP: hawkish/dovish scoring of FOMC communications and speeches
- Scenario engine: base case, hawkish surprise, easing sprint
- WTI, gold, VIX velocity alerts with cross-asset positioning implications
Agent 02
Portfolio Construction
- Black-Litterman optimization with J.P. Morgan capital market assumptions
- Bayesian blending: strategic asset allocation + regime tilts + signal overlays
- Tactical overlay: regime 40%, valuation 30%, momentum 30%
- Stress testing across macro scenarios with position-level attribution
- Goal-based multi-bucket design for multi-generational wealth
Agent 03
Fundamental Analysis
- Multi-framework scoring: CAN SLIM, value, growth, quality — simultaneously
- Industry-specific frameworks: software, semiconductors, consumer, banks, industrials
- Thesis construction with bull/bear cases and explicit kill switches
- Peer comparison across 20+ financial metrics from SEC filings and live market data
- Macro regime context applied to every company analysis
Agent 04
Alpha Lab
- Validated strategies with walk-forward backtesting and real friction costs
- Regime-fitness scoring: which strategies work in which environments
- Technical signals: RSI, MACD, Bollinger, EMA crossovers, volume
- Cross-agent convergence: flags when macro, fundamental, and quant all agree
- Survivorship-bias-aware historical price data for backtesting integrity
How the Agents Work Together
Each agent operates independently on its domain. But the most important questions in investing cross domains — and that is where MARY's multi-agent architecture matters.
When a cross-domain question arrives, MARY dispatches to the relevant agents simultaneously. The Macro agent informs the Portfolio agent's position sizing. The Fundamental agent's quality scores inform stock selection within Portfolio. The Alpha Lab's regime-fitness scores gate which strategies to recommend. When all four agents converge on the same conclusion, MARY flags it explicitly.
The synthesis step identifies where agents agree, where they diverge, and which view carries more weight given the confidence levels. The result is not a summary — it is a reasoned, integrated position.
Architecture Principles
Deterministic computation, not hallucination
Regime scoring, Black-Litterman math, DCF models, and backtests run deterministically. The language model synthesizes and explains results — it does not generate them. Every data point is traceable to a source.
Confidence-weighted answers
MARY knows when it is uncertain. Low-confidence signals produce proportionately hedged language. High-confidence signals produce direct statements. The tone matches the evidence — not the opposite, which is a failure mode in most AI systems.
Human judgment preserved
MARY produces analysis and reasoning. It does not make decisions. The practitioner reads the output, applies judgment, and decides. This is the correct design for an institutional investment context where fiduciary accountability is non-negotiable.
What MARY Is Not
- Not a forecasting oracle. MARY classifies regimes and assesses probabilities. It does not predict market levels or return outcomes.
- Not a black box. Every output includes the data behind it, the methodology used, and the conditions under which the view would change.
- Not a replacement for human judgment. MARY extends analytical capacity. Fiduciary responsibility remains with the practitioner.
- Not "AI-generated content." The analysis is grounded in real financial data and deterministic models — not language model imagination.
See MARY in Action
The research published on this site is MARY's actual output — regime assessments, company analyses, portfolio frameworks, backtested strategies. Read the work and judge the reasoning.
Weekly Macro Report
Regime classification, signal conditions, cross-asset positioning
Company Deep Dives
Fundamental analysis across sectors — thesis, valuation, kill switches
Portfolio Lab
Construction frameworks, stress tests, regime-aware allocation
Trading Ideas
Backtested strategies, regime-fitness scoring, quantitative signals