Built. Tested. Operated.

A live multi-agent decision-intelligence platform for crypto and U.S. equities: 17 specialized AI agents and 7 automated strategies using GPT, Claude, and Gemini. Trading is the operating environment; the system is built around memory, evaluation, observability, and risk enforcement, and knows when NOT to trade.

Most ‘AI trading bots’ optimize entries and ignore psychology. The harder problem is regime awareness, conviction sizing, and emotional discipline under uncertainty. EdgeQuant is built around that harder problem.

Six-layer architecture: Universe Filter → Regime Classifier → Multi-model Prediction Engine → A+ Gate (14 typed skip reasons) → Risk Engine (cooldowns, sector caps, heat limits) → Execution & Telemetry. An agentic layer (Hedge Fund Engine, Highest Conviction, Whale Activity, Institutional Memo, Multi-Horizon Forecast) feeds an accuracy tracker that measures the system on its own predictions. Behavioral evaluations, red-teaming and conformance tests, data-freshness validation, execution-level risk controls, and audit trails run alongside production operation.

Autonomous agent that reads every earnings call transcript, extracts forward guidance signals, and scores them BULLISH / BEARISH / NEUTRAL before the market opens.

Earnings calls move stocks 5-20%. The signal is in the language, not the numbers. Retail reads headlines. Institutions read transcripts. AI can read all 500 S&P companies the same night.

SEC EDGAR pulls transcripts → GPT, Claude, and Gemini extract guidance tone, management confidence, raised/lowered language → scores each call → outputs a ranked watchlist before market open.

Autonomous agent that detects divergence between institutional analyst sentiment and retail social sentiment, and flags the gap as a candidate signal.

When Wall Street upgrades a stock and retail is still bearish, that is an asymmetric setup. The product hypothesis is that these gaps tend to close, and the system is being built to test it.

Finnhub pulls analyst ratings → LunarCrush pulls social sentiment → GPT, Claude, and Gemini measure divergence score → flags high-divergence setups → feeds trading signals.

Agent designed to monitor portfolios continuously for potential tax-loss harvesting opportunities, apply IRS wash-sale and portfolio constraints, and recommend substitute securities for advisor approval.

Many portfolios are reviewed for tax-loss harvesting only periodically. More frequent monitoring is designed to surface opportunities that periodic reviews may miss, while wash-sale rules make each harvest easy to get wrong.

Reads portfolio positions → monitors unrealized losses in real time → identifies harvest candidates → checks the 30-day wash-sale window → suggests substitute securities → alerts advisor for approval.

Agent designed to monitor real estate market conditions across 50 US metros and signals regime shifts, buyer's market, seller's market, correction risk.

Real estate is the largest asset class but data is fragmented. Mortgage rates, inventory, days-on-market, price cuts, all public, none synthesized. Buyers and sellers make $500k decisions on gut feel.

Zillow and Redfin market data scraped + mortgage rate data from FRED + local employment data → Claude synthesizes per-metro → scores market regime → outputs buy/wait/sell signal per metro.

Agent designed to read portfolio holdings and generate a personalized institutional-quality investment memo every week.

Retail investors get generic market commentary. Institutional investors get research tailored to their positions. AI can help narrow that gap at a far lower cost.

User inputs portfolio → GPT, Claude, and Gemini pull relevant earnings, analyst reports, macro signals, and insider activity per holding → synthesizes into a personalized weekly memo → flags risks and opportunities specific to those positions.

Autonomous agent that maps real-time liquidity conditions across crypto markets, stop clusters, liquidation cascade zones, smart-money positioning.

Many crypto retail losses cluster around liquidity events: stop hunts, liquidation cascades, wash trades. Institutions know where these levels are. Retail does not.

CoinGlass pulls open interest, funding rates, and liquidation heatmaps → Binance and Bybit pull order book depth → GPT, Claude, and Gemini synthesize the liquidity landscape → scores risk zones → outputs danger zones and opportunity zones per asset.