neXaQuant

Democratizing quantitative trading analysis with multi-agent AI, turning anyone into a data-driven trader. Backtest any strategy in seconds and verify influencer claims instantly.

Comment

Inspiration

Losing $10,000 while trying to learn how to invest in the markets and blindly following so-called "gurus" on social media who claimed to be making 10-20% returns daily! We realized there was no easy way to validate these claims or backtest trading strategies without coding expertise. neXaQuant was born from the need to democratize quantitative trading analysis and hold financial influencers accountable.

What it does

neXaQuant is an AI-powered trading intelligence platform that transforms natural language into actionable trading insights:

  • Strategy Builder: Describe any trading strategy in plain English, and our multi-agent AI system parses it, backtests it against historical data, and shows you real performance metrics (Sharpe ratio, CAGR, max drawdown)
  • Guru Analyzer: Paste transcripts from trading influencers or generate sample calls, and we score their performance on a 0-100 scale based on profitability, risk-adjusted returns, consistency, and risk management
  • Agent Marketplace: Autonomous AI buyer and seller agents negotiate strategy purchases through multi-round strategic reasoning, demonstrating cutting-edge agentic decision-making
  • NFT Minting: Mint your backtested strategies as tradeable NFTs on Solana with full metadata (performance metrics, code, parameters)
  • Market Intelligence: Real-time technical analysis with SMAs, volatility, trend detection, and comprehensive market insights

Every feature shows you exactly what your portfolio would be worth if you started with $1,000.

How we built it

Tech Stack:

  • Frontend: React + Vite + Tailwind CSS with Recharts for data visualization
  • Backend: FastAPI with custom pandas-based backtesting engine
  • AI Layer: OpenRouter for multi-model intelligence (GPT-4o-mini, Claude, Llama) - different models for different tasks
  • Blockchain: Solana for strategy NFT minting with on-chain metadata
  • Data: Yahoo Finance API for real historical OHLC data
  • Deployment: Vercel (frontend) + Render (backend)

Architecture Highlights:

  • Multi-agent system with specialized agents (Parsing, Backtest, Risk Scoring, Guru Analysis, Marketplace Negotiation)
  • Buyer agents with budget constraints negotiate with multiple Seller agents (aggressive, balanced, cooperative personalities)
  • Each agent uses LLM-powered strategic reasoning with memory of negotiation history
  • Comprehensive 4-component guru scoring: Profitability (40%), Risk-Adjusted Returns (30%), Consistency (20%), Risk Management (10%)

Challenges we ran into

  1. Agent Negotiation Complexity: Building buyer/seller agents that could reason strategically and negotiate across multiple rounds while maintaining conversation context was challenging. We solved this with turn-based prompting and explicit negotiation history tracking.

  2. Backtesting Accuracy: Creating a reliable backtesting engine that handles edge cases (missing data, split-adjusted prices, signal timing) while remaining simple enough for AI to generate strategies for. We built a custom pandas-based engine with robust error handling.

  3. Deployment Headaches: Railway, our initial choice for backend hosting, kept failing with cryptic errors. We pivoted to Render.com and got it deployed within minutes - a lesson in choosing the right tools!

  4. CORS Configuration: Connecting our Vercel frontend to our Render backend involved multiple CORS debugging sessions and environment variable updates across deployments.

  5. Real-time Performance: Ensuring the guru analyzer could process multiple trades and calculate portfolio-level metrics efficiently without timeout issues.

Accomplishments that we're proud of

  • Multi-Agent AI Marketplace: Built a fully functional agent negotiation system where autonomous agents reason, strategize, and complete deals - pushing the boundaries of agentic AI
  • Comprehensive Guru Scoring: Created a rigorous 4-component scoring methodology that tracks actual portfolio performance, not just win rates
  • Natural Language Trading: Made quantitative strategy testing accessible to anyone - no coding required
  • Beautiful UX: Crafted smooth animations, intuitive workflows, and data visualizations that make complex financial metrics understandable
  • Full-Stack Integration: Successfully connected AI agents, blockchain, financial APIs, and real-time data into one cohesive platform
  • 48-Hour Build: Went from idea to fully deployed production app on nexaquant.tech in one hackathon weekend!

What we learned

  • Multi-Model Orchestration: Different AI models excel at different tasks - GPT for reasoning, Llama for code generation, Claude for summarization
  • Agent Design Patterns: Building autonomous agents requires careful prompt engineering, explicit state management, and strategic fallback logic
  • Financial Math: Deep dive into Sharpe ratios, drawdown calculations, annualized volatility, and risk-adjusted return metrics
  • Deployment Strategy: Sometimes the "easier" solution (Render over Railway) saves hours of debugging
  • User-Centric Design: Financial tools don't have to be boring - animations and clear visualizations make complex data accessible

What's next for neXaQuant

  • Live Trading Integration: Connect to brokerage APIs for paper trading and eventually live execution
  • Social Features: Share strategies, follow top performers, create a community leaderboard
  • Advanced Strategies: Support for options, futures, multi-asset portfolios, and machine learning-based strategies
  • Enhanced NFTs: Make strategy NFTs actually executable - buy an NFT, run the strategy automatically
  • Mobile App: Take neXaQuant on-the-go with iOS and Android apps
  • Institutional Features: Portfolio optimization, risk management tools, and compliance reporting for professional traders
  • Mainnet Launch: Move from Solana devnet to mainnet for real trading strategy marketplace
  • AI Strategy Generator: Let AI create novel strategies based on market conditions and user risk preferences

neXaQuant isn't just a hackathon project - it's the foundation for democratizing quantitative trading and bringing transparency to financial social media. No more blindly following gurus. No more $10,000 losses. Just data-driven decisions. 📈

Built With

Share this project:

Updates