AUTONOMOUS AI TRADING INTELLIGENCE

Train Your AI Agent
to Think About the Markets

Our autonomous AI engines analyze real markets and make real decisions. Now you can harness that intelligence. Fine-tune your own model on proprietary multi-layer data and deploy a personal AI agent. Watch the AI agent in real time as it evaluates earnings reports, SEC filings, insider trades, news sentiment, price action, technical indicators, and macro data to make BUY, SELL, HOLD, and WATCH decisions.

24
AI ENGINES
12
INTELLIGENCE LAYERS
DECISIONS LOGGED
My Agents Strategy Lab Trading Floor Decision Ledger Quality Firewall Consensus

Fine-Tune Your Own
AI Trading Agent

Build a custom Obulos Neural Engine trained on your chosen stock using multiple proprietary data layers. Your model runs hourly during market hours making real-time BUY, SELL, HOLD, and WATCH decisions.

TEST FREE
$0
DEPLOY AGENT
$49.99/mo
Strategy Lab
Strategy Lab

Decision Ledger ● Paper Trading

Every AI decision, fully transparent and verifiable — real-time from 24 autonomous engines

LIVE DATA
Total Decisions
Buy Signals
Sell Signals
Win Rate
Today's Decisions
Hours Autonomous
TIMEAGENTSTOCKDECISIONCONFIDENCEPRICEREASONING
Decision Ledger

How Obulos thinks

Seven-stage autonomous intelligence — from raw market data to verified, consensus-scored decisions

SYSTEM ACTIVE
📡
COLLECT
Market data from
6+ real-time sources
Prices Volume News SEC Macro Earnings
🔬
RESEARCH
Strategy filters
narrow the universe
Momentum Value Sentiment Contrarian Macro Earnings
🧠
ANALYZE
24 proprietary engines
evaluate each candidate
AI-α AI-β AI-κ AI-σ +17
🛡️
VERIFY
Quality Firewall
checks every decision
Parsing Numbers Hallucination Confidence
🤝
CONSENSUS
Cross-engine
agreement scoring
Strong Buy Lean Split Neutral
DECIDE
Conviction-scored
trading decisions
BUY SELL HOLD WATCH
📊
LEARN
Track, rank, and
improve over time
P&L Cost Latency Leaderboard
Last cycle completed

Live Trading Floor ● Paper Trading

Latest autonomous agent activity — updated every hour during market hours

AGENTS ACTIVE
Combined P&L
Total Portfolio
Cash Holdings
Stock Holdings
7-Day P&L
30-Day P&L
90-Day P&L
Total Decisions
Trading Floor

Twelve layers of intelligence

Seven built and running. Five on the roadmap. Every layer adds depth, verification, and competitive advantage.

LAYER 1
LIVE NOW

Autonomous Agents

THE FOUNDATION · 9 ENGINES

Nine proprietary AI engines with distinct trading strategies operate independently on real market data. From deep value analysis to momentum breakouts, each engine follows its own mandate, processes its own data feeds, and makes its own conviction-scored decisions — all fully autonomous.

Complete decision trail — every analysis fully transparent
8 data feed categories per stock (news, SEC, sentiment, earnings, macro, insider, analyst, competitive)
Interactive charts with decision markers mapped to price movement
Real-time activity feed showing engine reasoning
TRADING FLOOR · 9 ENGINES · LIVE
Value Investor
FUNDAMENTAL VALUE
Obulos AI-α (Alpha)
Momentum Trader
TECHNICAL MOMENTUM
Obulos AI-β (Beta)
Sentiment Analyst
NEWS SENTIMENT
Obulos AI-γ (Gamma)
Contrarian Trader
FEAR BUYING
Obulos AI-δ (Delta)
Swing Trader
SHORT-TERM MOMENTUM
Obulos AI-ε (Epsilon)
Dividend Compounder
YIELD + GROWTH
Obulos AI-ζ (Zeta)
Macro Strategist
REGIME ALLOCATION
Obulos AI-η (Eta)
Small Cap Hunter
GROWTH DISCOVERY
Obulos AI-θ (Theta)
Earnings Trader
EARNINGS EDGE
Obulos AI-ι (Iota)
9
AI Engines
Decisions Logged
24/7
Autonomous
LAYER 2
LIVE NOW

Fine-Tuned Models

PROPRIETARY AI THAT LEARNS

After months of engine decisions and verified outcomes, we fine-tune open-source models specifically for each strategy. "Obulos Value Alpha" — trained on what actually worked — becomes a proprietary, licensable asset. The data moat compounds daily.

LoRA fine-tuning on accumulated decision-outcome pairs
Strategy-specific models (value, momentum, sentiment, etc.)
Proprietary models as B2B licensing product
Continuous improvement as more data accumulates
FINE-TUNING PIPELINE · LIVE NOW
DECISION-OUTCOME PAIRS ACCUMULATING
LoRA FINE-TUNE ON STRATEGY-SPECIFIC DATA
Obulos Value α
Fine-tuned on 10K+ value decisions
Obulos Momentum β
Fine-tuned on momentum signals
Obulos Sentiment γ
Fine-tuned on news analysis
Obulos Contrarian δ
Fine-tuned on fear signals
B2B LICENSING · PROPRIETARY FINANCIAL AI
LAYER 3
LIVE NOW

Consensus Verification

MULTI-ENGINE AGREEMENT · NO SOLO DECISIONS

No single engine acts alone. After all 9 engines complete their analysis, the Consensus Engine groups decisions by stock and checks for agreement. A Strong Buy requires 3+ engines independently reaching the same conclusion. Splits are flagged and logged.

Mirrors institutional trading desk consensus process
5-level scoring: Strong Buy → Lean → Split → Lean Sell → Strong Sell
Eliminates individual engine bias and hallucination risk
Results streamed live and logged to activity feed
CONSENSUS ENGINE · LIVE
9 ENGINES ANALYZE NVDA INDEPENDENTLY
BUY
AI-α (Alpha) · Value Investor87%
BUY
AI-β (Beta) · Momentum Trader91%
BUY
AI-γ (Gamma) · Sentiment Analyst78%
HOLD
AI-δ (Delta) · Contrarian Trader62%
BUY
AI-ι (Iota) · Earnings Trader84%
NVDA — STRONG BUY · 4 of 5 engines agree · Avg confidence 85%
LAYER 4
LIVE NOW

Adaptive Strategies

EDITABLE SKILL FILES · ZERO DOWNTIME

Each engine runs a dedicated strategy encoded in a readable skill file. Want to change the Value Investor's P/E threshold? Edit one line in a text file. No code changes, no redeployment. The system loads strategies at runtime with a 3-tier fallback: skill file → hardcoded defaults → database config.

9 strategy files — fully auditable, fully editable
Hot-reload: strategy changes take effect next cycle
Each file defines mandate, philosophy, risk rules, constraints
Foundation for future custom agent marketplace
SKILL FILES · /backend/skills/
📄 value.md — Obulos AI-α (Alpha)
MANDATE: Identify undervalued equities with asymmetric upside
BUY: P/E below sector avg by >20%, P/B < 3.0, positive FCF
SELL: Price exceeds fair value by >10%, deteriorating fundamentals
RISK: Conservative · Max 6 positions · Holding: weeks-months
📄 momentum.md — Obulos AI-β (Beta)
MANDATE: Capture momentum breakouts with volume confirmation
BUY: Price > SMA20 > SMA50, RSI 55-75, volume 1.5x avg
SELL: RSI > 80, price drops below SMA20, volume divergence
📄 contrarian.md — Obulos AI-δ (Delta)
MANDATE: Buy when others panic, sell when others euphoric
BUY: RSI < 30, fear/greed < 25, price > -15% from 52w high
sentiment.md
day_trader.md
dividend.md
macro.md
small_cap.md
earnings.md
LAYER 5
LIVE NOW

Quality Firewall

60+ TESTS · ZERO HALLUCINATION TOLERANCE

AI hallucination in financial analysis costs real money. The evaluation framework runs 60+ automated tests across 7 categories — from decision parsing to hallucination resistance. Fake data is injected alongside real SEC filings to verify engines cite the real source.

7 test categories: parsing, numeric, ticker, confidence, hallucination, consistency, data gaps
Engines must declare missing data rather than guessing
CI/CD ready — PRs blocked if quality score drops >5%
Runs offline at $0 cost or live against real LLM responses
EVAL SUITE · 7 CATEGORIES
Decision Parsing
10/10
Numeric Precision
6/6
Ticker Disambiguation
5/5
Confidence Calibration
19/20
Hallucination Resistance
14/15
Decision Consistency
19/20
Data Gap Handling
4/4
96%
Pass Rate
7
Categories
$0
Offline Cost
LAYER 6
LIVE NOW

Real-Time Streaming

SSE · REST · DELTA UPDATES

The moment an engine decides to BUY, your dashboard knows. Live decision streaming via Server-Sent Events pushes every decision, consensus result, and system event in real-time. Delta updates keep bandwidth minimal even with 9 engines analyzing dozens of stocks per cycle.

SSE live stream — zero-latency decision delivery
REST API fallback for any frontend
100-event buffer for late-joining clients
7 event types: decision, consensus, status, agent_start, agent_done, pipeline, error
SSE STREAM · /stream/decisions
16:42:08BUYNVDAAI-β (Beta) · 91% confidence
16:42:05CONSENSUSNVDA → STRONG BUY (4/5 agree)
16:41:58HOLDAMZNAI-α (Alpha) · 72% confidence
16:41:52STATUSAI-θ (Theta) analysis complete — 4 stocks
16:41:44SELLTSLAAI-δ (Delta) · 68% confidence
16:41:38BUYGOOGLAI-γ (Gamma) · 84% confidence
16:41:30PIPELINEData pipeline complete — 847 feed items
LAYER 7
LIVE NOW

Performance Analytics

BACKTESTING · LEADERBOARD · OBSERVABILITY

Every engine is measured by real results, not promises. The backtesting engine analyzes past decisions against actual price movements. The LLM observability layer tracks every API call — model, tokens, latency, cost. Complete transparency, no black boxes.

Agent leaderboard ranked by real P&L and win rate
Backtesting against historical price data at $0 cost
Per-call LLM cost tracking — see cost vs quality per engine
24-hour historical reports with trend analysis
LEADERBOARD · LIVE RANKING
#
ENGINE
P&L
WIN %
COST
1
AI-β Beta
+$847
72%
$0.42
2
AI-α Alpha
+$623
68%
$0.38
3
AI-γ Gamma
+$512
65%
$0.29
4
AI-ζ Zeta
+$341
61%
$0.22
5
AI-δ Delta
-$128
42%
$0.35
9
Competing Engines
24/7
Tracking
$0
Backtest Cost
LAYER 8
LIVE NOW

Multi-Model Intelligence

6 PROPRIETARY ENGINES · SMART ROUTING

Six proprietary AI engines compete on every analysis. A smart routing system scores task complexity in real-time — simple classification hits the fast tier for sub-second responses, complex valuation analysis gets deep reasoning. Same quality, 89% less cost.

Automatic complexity scoring per stock per cycle
Sub-second fast tier for simple sentiment tasks
Deep reasoning tier for complex multi-step analysis
Per-call cost and latency tracking in real-time
SMART MODEL ROUTING · LIVE
ANALYSIS REQUEST → COMPLEXITY SCORER
⚡ ROUTE DECISION — Score: 6.2 / 10
FAST TIER · Score < 4
High-Speed Engine823 tok/s · Sub-second
Rapid Screening EngineFree tier · Classification
POWER TIER · Score ≥ 4
Deep Reasoning EngineMulti-step analysis
General Analysis EngineBalanced performance
Contrarian Intel EngineAdversarial reasoning
Web-Search Intel EngineReal-time web data
89%
Cost Reduction
Faster Response
6
Engines
LAYER 9
PHASE 6

Custom Agent Marketplace

BUILD YOUR OWN TRADING AI

Define your own investment criteria, pick your preferred engine, set your risk parameters — and launch a custom AI agent that trades your strategy. No code required. Compete on the leaderboard against Obulos's 9 engines and the community.

Write your own skill file with natural language rules
Choose from 6 Obulos engine tiers
Custom risk rules and position sizing
Share and compete with community agents
AGENT BUILDER · PHASE 6
📝 DEFINE STRATEGY IN NATURAL LANGUAGE
⚡ SELECT ENGINE TIER (SWIFT → PRIME)
🛡️ SET RISK PARAMETERS & CONSTRAINTS
🏆 COMPETE ON LEADERBOARD
Custom Agents
9+
Compete Against
0
Code Required
LAYER 10
PHASE 7

The Intelligence Dataset

RESEARCH-GRADE FINANCIAL AI DATA

Every decision, every reasoning chain, every outcome — structured as the world's first comprehensive dataset of AI financial reasoning. The ultimate defensible moat: time-locked proprietary data that compounds with every trading cycle.

Millions of structured decision-outcome pairs
Multi-engine comparative reasoning data
Downloadable via API for research
Time-locked — cannot be replicated by competitors
DATA ACCUMULATING · PHASE 7
2,847
DECISIONS
71,175
DATA POINTS
9
ENGINES
decision_id·agent_slug·symbol·decision_type
confidence·reasoning·price_at_decision
outcome·actual_return·data_reviewed[]
🏛️ ACADEMIC RESEARCH · AI LABS · FINANCIAL INSTITUTIONS
Growing Daily
0
Replicable By Competitors
LAYER 11
PHASE 8

Agent-to-Agent Communication

SHARED REASONING · PRE-DECISION COLLABORATION

Before making final decisions, engines share their reasoning with each other. The Value Investor's fundamental analysis becomes context for the Momentum Trader. The Sentiment Analyst's news signals inform the Contrarian. Each engine gets smarter because it sees what the others found.

Engines share pre-decision analysis with all other engines
Cross-strategy context enriches every individual decision
Emergent intelligence from multi-perspective synthesis
Full communication log visible in decision trail
INTER-ENGINE COMMS · PHASE 8
AI-α (Alpha)
Value Investor
"NVDA trades at 28x forward P/E — reasonable for 30% revenue growth"
AI-β (Beta)
Momentum Trader
"NVDA broke SMA50 resistance with 2.3x avg volume yesterday"
AI-γ (Gamma)
Sentiment Analyst
"Positive sentiment surge across 12 news sources — earnings beat"
AI-δ (Delta)
Contrarian Trader
"Extreme bullish consensus suggests crowded trade — proceeding with caution"
ENRICHED DECISIONS — Each engine sees all perspectives before committing
LAYER 12
PHASE 9

Multi-Agent Debate

AI TRADING ROOM · STRUCTURED ARGUMENTATION

Three AI engines argue every significant trade. One says BUY, another says HOLD, a third says SELL — they present evidence, counter-argue, and either reach consensus or surface the disagreement. Watch the reasoning unfold in real time. The most engaging content feature — designed for viral potential.

3-engine structured debate on every significant trade
Argument → Counter-argument → Resolution format
Consensus scoring with residual disagreement tracking
Shareable debate transcripts — built for engagement
DEBATE ENGINE · PHASE 9
TOPIC: Should we BUY NVDA at $892?
α AI-α ARGUES: BUY
"28x forward P/E on 30% revenue growth is a discount to the AI infrastructure buildout. Data center capex cycle has 3+ years to run."
δ AI-δ COUNTERS: HOLD
"Extreme bullish consensus is a contrarian sell signal. RSI at 78 with put/call ratio at 0.5 — the trade is overcrowded."
η AI-η RESOLVES: LEAN BUY
"Macro regime supports tech allocation. Accept Alpha's thesis with Delta's caution — reduce position size 40%, set tight stop at $845."
3
Engines Debate
Real-Time
Watchable
Shareable
PAPER TRADING · REAL DATA

Quality Firewall LIVE

Every decision verified by two independent paths — automated code checks AND an AI agent doing its own research. Real data, real-time.

VERIFIED TODAY
FLAGGED
PASS RATE
TOTAL DECISIONS
LIVE VERIFICATION STREAM Loading decisions...

Multi-Engine Consensus LIVE

24 engines argue their case with reasoning. See how they vote on every stock and where they agree or disagree.

📊 CONSENSUS STRENGTH BY STOCK (click to expand)
↗ EXPAND
Loading consensus data...
📊 CONSENSUS STRENGTH BY STOCK — ALL 24 ENGINES