APEX — Autonomous Predictive Exchange Organism

Vercel
application badge
Created by team APEX Trading Intelligence on March 31, 2026
ERC-8004Kraken - Trading Performance (PnL)Kraken - Social Engagement

APEX (Autonomous Predictive Exchange) is the world's first trustless, self-learning, multi-agent AI trading organism. Traditional AI trading agents are black boxes — no audit trail, no accountability, no verifiable decision chain. APEX solves this with 8 specialized AI agents coordinated under strict separation of concerns, with every trade decision cryptographically signed via EIP-712 and permanently recorded on Ethereum. The system runs a 60-second trading cycle: DR. YUKI TANAKA fetches live BTC/USD prices from Kraken; DR. JABARI MENSAH analyzes 40 crypto news articles via Azure GPT-4o for sentiment; DR. SIPHO NKOSI enforces 8 risk guardrails including position limits, drawdown caps, and circuit breakers; DR. ZARA OKAFOR (OpenRouter/Qwen3-72B) makes the final BUY/SELL/HOLD decision; DR. PRIYA NAIR submits an EIP-712 signed trade intent to our RiskRouter smart contract; a validation checkpoint is posted to our ValidationRegistry; ENGR. MARCUS ODUYA executes via Kraken; and DR. LIN QIANRU updates the PPO reinforcement learning policy. The result: 1,859 on-chain trade proofs, validation score 98/100, reputation score 95/100 (ERC-8004 standard), Sharpe ratio 1.84, and leaderboard rank #5 of 67 teams — all verifiable on Ethereum Sepolia by anyone, at any time. APEX targets institutional traders, DeFi protocols, and prop trading firms who need provable AI decision-making — not just performance claims. The ERC-8004 identity and reputation system creates a permanent, trustless trading record that can bootstrap real institutional capital allocation.

Category tags:

"Application of Technology 5/5 8 specialized agents with named personas, clear separation of concerns. PPO reinforcement learning for policy updates. EIP-712 signed intents. RiskRouter + ValidationRegistry dual on-chain checkpoint. 1,030+ verifiable on-chain intents. Azure Container Apps production infra. Business Value 4.5/5 "Trustless trading record for institutional capital" is a strong pitch. The dual-chain (RiskRouter TX + ValidationRegistry TX) every trade is compelling. Live Risk API publicly queryable. Target: institutional DeFi protocols + prop trading. Presentation 4.5/5 Named agents with roles + credentials is memorable. "You don't trust the developer. You trust the math." is a great tagline. README is clean and verifiable. Dashboard exists. Competing for 4 prize categories shows ambition. Originality 4.5/5 8-agent named persona architecture is distinctive. PPO RL for policy updates is a genuine ML contribution. The separation-of-concerns model (Priya signs, Sipho checks, Lin learns) is a strong design pattern. Standout elements: ✅ 8 specialized agents with distinct models (OpenRouter/Qwen3-72B, Azure GPT-4o, SambaNova/Qwen2.5, Google Gemini-2.5-Pro, Groq/Llama3.3-70B) ✅ PPO reinforcement learning updating trading policy based on outcomes ✅ 1,030+ on-chain trade intents (verified on Sepolia Etherscan) ✅ Agent ID #26 on ERC-8004 ✅ Live public Risk API: apex-risk-api.ambitiouspebble-9b46cb11.francecentral.azurecontainerapps.io/risk/status ✅ Dual on-chain checkpoint (RiskRouter + ValidationRegistry) for every trade ✅ Validation Score 97/100, Reputation Score 93/100 ✅ Sharpe ratio 1.84 ✅ Rank #6 of 67 teams ✅ Azure production infrastructure, paper→live in 2 config changes ⚠️ Claims $10k Best Trustless prize + $2.5k Risk Guardrails + $2.5k Validation + $1.8k Kraken Performance = $16.8k potential (multi-category submissions) Total: ~4.75/5 — Competing directly with Swiftward for the top tier. Very different architecture (named agent personas + PPO RL + institutional focus) but equally impressive execution. "

avatar

Sanem Avcil