Nexus Investing
|
Autonomous AI-Powered Trading System
An autonomous trading engine powered by 8 AI agents across 5 LLM families, managing 6 trading profiles across 6 brokerage accounts — built to think, debate, and trade like a hedge fund.
57×
Backtest Return (7yr)
8
AI Agents
6
Trading Profiles
38,284
Backtests Validated
5
LLM Families
0%
Ruin Probability
Architecture
Two-Lane Engine Design
A Fast Lane for intraday momentum with sub-second Python execution, and a Strategic Lane for position trading with full multi-agent AI council deliberation.
⚡ FAST LANE
Day Trading Engines
Four specialized engines for intraday momentum, options, short selling, micro-cap pumps, and small-cap breakouts. All positions flattened by EOD.
APEX Elite
Aggressive
Pump & Dump Hunter
Ross Cameron
- 🎯 Python gatekeeper for instant strategy selection (<10ms)
- 📈 Open universe scanning — 65+ candidates per cycle via FMP
- 💡 6-detector composite scoring for pump detection
- 🧠 LLM catalyst grading with 4-provider fallback chain
- 🔴 Redis Streams signal bus — <5ms emergency latency
- 📊 ATR trailing stops let winners run (2.0× multiplier)
🧠 STRATEGIC LANE
Position Trading Engine
Multi-agent AI council deliberation for longer-term positions. Quality-momentum screening, dividend aristocrats, and covered call income.
Moderate
Conservative
- 👥 8 AI agents debate every trade decision
- 📈 Bull vs Bear researcher adversarial analysis
- 💰 Quality-momentum stock screening (ROE >15%, D/E <0.5)
- 🏆 Dividend Aristocrat filtering (25+ year streaks)
- 📋 Covered call income overlay on held positions
- 🛡 Post-earnings drift capture (PEAD strategy)
AI Council
The AI Council
Every trade decision is evaluated by 8 specialized agents across 5 LLM families. No single model controls the outcome — diverse reasoning ensures robust decisions.
Market Analyst
GPT-4o
Price, volume & technicals
Fundamentals
Claude Sonnet
Revenue, margins, valuation
News & Sentiment
Gemini 2.5 Pro
Breaking news, social pulse
Calendar Agent
DeepSeek R1
Earnings, events, catalysts
Bull Researcher
Claude Opus 4.6
Argues FOR the trade
Bear Researcher
GPT-4o
Argues AGAINST the trade
Risk Manager
Claude Sonnet
Portfolio constraints & sizing
Trader (CIO)
Claude Opus 4.6
Final decision authority
Profiles
Six Trading Profiles
From conservative decade-long wealth building to ultra-aggressive micro-cap pump detection — each profile runs its own engine, risk model, and brokerage account.
💎
APEX Elite
Multi-Strategy Rotation • Python Gatekeeper • Backtest-Validated
57×
7yr Return
78%
CAGR
1.27
Sharpe
7/7
Win Years
🔥
Aggressive
Momentum Breakouts • Graduated Regime Gate • Kill Zones
Day
Hold Period
5
Max Positions
A/B
Catalyst Grade
3%
Daily Loss Cap
💣
Pump & Dump Hunter
6-Detector Scoring • Cash-Governed Entry • Drop-to-Upgrade
6
Detectors
25/35%
Exposure Tiers
85+
Composite Entry
10%
Daily Loss Cap
⚡
Ross Cameron
Small-Cap Momentum • First Hour Only • Pattern Detection
3
Max Trades/Day
60m
Entry Window
LONG
Direction
6%
Daily Loss Cap
⚖️
Moderate
Quality-Momentum • Core + Satellite • Covered Calls
12
Max Positions
Days
Hold Period
65/35
Core / Satellite
8%
Max Per Stock
🏙️
Conservative
Dividend Aristocrats • Capital Preservation • 25+ Year Streaks
15
Max Positions
Years
Hold Period
85/15
Core / Satellite
5%
Max Per Stock
Fast Lane
APEX Elite Engine
The crown jewel of the Fast Lane — APEX runs a multi-strategy state machine where a Python gatekeeper selects the optimal strategy in <10ms based on regime, time window, and VIX. Parameters validated across 38,284 backtests and 10,371 trades.
6 Adaptive Strategies
APEX selects from 6 strategies via a data-driven condition map built from 1.69 million analyzed trades. 3 strategies are currently active (proven edge), 3 are disabled (unprofitable in out-of-sample testing).
APEX ELITE
VWAP Mean Reversion
Lunch Range Reversion
Power Hour Continuation
Gap & Go
Opening Range Breakout
Breakdown Short
Analytics
Institutional-Grade Analytics
The system ingests data that most retail investors never see — dark pool prints, Level 3 options flow, institutional order flow, and social sentiment — all scored and fed into the decision engine.
Dark Pool Detection
Detects large institutional block trades hidden from public exchanges. Unusual dark pool volume signals smart money positioning.
Options GEX + Max Pain
Gamma Exposure analysis reveals where market makers must hedge. Max pain identifies the price where most options expire worthless.
Confluence Scoring
Weighted multi-factor entry quality scoring. Combines technicals, flow, volume, and catalyst strength into a single 0-100 confidence score.
6-Detector Pump Scoring
Volume spike, price velocity, social mentions, unusual options, low float, and SEC filing signals. Composite 0-100 score with detector-level weighting.
Unusual Whales + Social
Real-time dark pool alerts, options flow discovery, StockTwits buzz, Finnhub news, Quiver Quant WSB tracking, and Twitter pump-language detection.
Counterfactual Tracking
Records every blocked signal with entry price, then tracks hypothetical outcomes at 30, 60, 120 minutes and EOD. Data-driven validation of whether blocks were correct.
Risk
Graduated Regime Detection
The single most important innovation — don't trade in bad markets. The graduated gate evaluates catalyst quality and confluence score to allow high-quality setups through in YELLOW at reduced size, while blocking everything in RED.
🌱
GREEN
VIX < 25 AND SPY above 50MA & 200MA
ALL STRATEGIES • 1.0× SIZE
⚠️
YELLOW
VIX 25-35 OR SPY below 50MA. Graduated gate: A-grade + confluence ≥70 = 0.75×, B-grade + ≥50 = 0.25×
QUALITY-GATED • 0.25-0.75×
🔴
RED
VIX > 35 OR SPY below 200MA
ALL BLOCKED
🚨
CIRCUIT BREAKER
Daily or weekly loss limit hit. Progressive risk reduction, then full halt.
EMERGENCY HALT
Performance
Backtest-Validated Performance
APEX parameters validated across a 60-configuration parameter sweep with 38,284 backtests and 10,371 trades. Scoring 5.0 out of 6 ROBUST on the walk-forward efficiency framework.
| Metric | Value | Notes |
|---|---|---|
| Starting Capital | $500 | Minimum viable starting capital |
| Final Equity (7yr) | $33,703 | 57× return — $500 to $33.7K |
| CAGR | 78% | 37% excluding 2022 outlier year |
| Sharpe Ratio | 1.27 | Above 1.0 = institutional grade |
| Profit Factor | 1.275 | Gross profit / gross loss |
| Win Rate | 55% | Across 10,371 validated trades |
| Profitable Years | 7 / 7 | 100% — every year profitable |
| Max Drawdown | -32% | Worst peak-to-trough decline |
| Walk-Forward Efficiency | 1.27 | >0.5 = robust, not overfit |
| Monte Carlo P46 | Positive | 46th percentile still profitable |
| Ruin Probability | 0% | Zero probability of account wipeout |
| Overall Score | 5.0 / 6 | ROBUST classification |
Learning
Self-Learning Feedback Loop
Every trade is analyzed, scored, and fed back into the system. The engine auto-tunes its own parameters through reinforcement learning and postmortem analysis.
Trade Logger
Real-time
→
Layered Memory
FinMem 3-tier
→
Postmortem
End of day
→
Reinforcement
VRL scoring
→
Optimization
Auto-tune
Comparison
What Sets Nexus Apart
Institutional-grade capabilities typically reserved for hedge funds, accessible to an individual investor.
Typical Retail Systems
- Single algorithm, one-size-fits-all
- No debate or second opinion
- Static rules, no learning
- No risk profile customization
- No position sizing by volatility
- Manual trade execution
- No dark pool or options flow data
Nexus Investing
- 8 specialized AI agents collaborate on every trade
- Bull vs Bear adversarial debate built in
- Self-learning feedback loop after every trade
- 6 distinct profiles with tailored strategies
- VIX-based dynamic sizing + graduated regime gate
- Fully autonomous execution across 6 accounts
- Dark pool, GEX, options flow, social sentiment
Infrastructure
Production Infrastructure
Multi-broker execution, real-time data feeds, fault-tolerant state management, and a Streamlit command dashboard.
🔗 Multi-Broker
Alpaca (US), Questrade (CAD), and IBKR support via broker abstraction layer. Profile-isolated credentials across 6 accounts.
3 broker adapters • 6 accounts
📈 Market Data
FMP primary data source with yFinance fallback. Theta Terminal for real-time options flow and dark pool data. Unusual Whales for flow discovery.
4+ data providers
🔒 Risk Controls
Circuit breaker with file-persisted state. Daily, weekly, and monthly loss limits. Progressive risk reduction on drawdowns. Per-profile isolation.
4-tier loss limits
💾 State & Signal Bus
Redis Streams for real-time inter-engine communication (<5ms latency). JSON file fallback for persistence. 30-second broker reconciliation loop.
Redis Streams + JSON fallback
📊 Trade Analytics
Institutional-grade metrics: Sharpe, Sortino, Calmar, Max Drawdown, Profit Factor, Expectancy. Equity curve charts, benchmark comparison (Alpha, Beta vs SPY).
CAPM-based benchmarking
🚀 Captain's View Dashboard
3×2 grid overview of all 6 accounts at a glance. Equity, daily P&L, positions, engine status. One-click navigation to any profile. Auto-refreshes every 30 seconds.
Streamlit • Dark Theme
Stack
Technology Stack
Built on proven technologies and industry-standard APIs.
Python
Core Engine
LangGraph
Agent Orchestration
Streamlit
Dashboard UI
Redis 7
Signal Bus
TimescaleDB
Time-Series DB
Docker
Containerization
FMP API
Market Data
Theta Data
Options / Dark Pool
Unusual Whales
Flow Discovery
Alpaca
Broker & Execution
Questrade
Broker (CA)
IBKR
Broker (Global)
GPT-4o
OpenAI
Claude Opus 4.6
Anthropic
Gemini 2.5 Pro
Google
DeepSeek R1
DeepSeek