TradeX

AI-powered trading platform with real-time market data, Gemini Flash trading signals, paper trading, and enterprise-grade security, built for every type of trader.

Comment

Inspiration

As retail traders, we found existing platforms either too complex for beginners or too limited for serious analysis. Bloomberg Terminal costs $24,000/year. Robinhood lacks depth. We wanted to build something in between — a platform that gives every trader access to professional-grade tools without the price tag or complexity.

What it does

TradeX is a full-stack trading platform that brings together:

  • Real-time market data across crypto, stocks, forex, and commodities
  • AI-powered trading signals using Google Gemini Flash — with BUY/SELL/HOLD recommendations, confidence scores, price targets, and plain-English explanations
  • Paper trading engine — practice with $100,000 virtual balance at live market prices
  • Portfolio management — track holdings, P&L, and asset allocation
  • Financial news aggregated in real time
  • Two-factor authentication and enterprise-grade security (2FA, CSRF, rate limiting, bcrypt)
  • Password reset via email with branded Resend transactional emails
  • 7-day persistent sessions that survive page refreshes

How we built it

Frontend: React 18 + Vite, React Router, Recharts for charts, Framer Motion for animations, CSS Modules for scoped styling.

Backend: Node.js + Express with a layered architecture — Routes → Controllers → Services → Repositories → PostgreSQL (Supabase).

AI: Google Gemini Flash 1.5 analyzes technical indicators (RSI, MACD, Bollinger Bands, EMA, SMA) and returns structured trading signals with full explainability. Rule-based analysis acts as fallback when Gemini is unavailable.

Auth: JWT tokens stored in localStorage with a 7-day expiry, sent as Authorization Bearer headers on every request — bypassing cross-domain cookie restrictions on Render's free tier.

Infrastructure: Frontend on Render Static Site, backend on Render Web Service, PostgreSQL on Supabase. Keep-alive pings prevent cold starts. Retry logic handles slow server wakeups.

Challenges we ran into

  • Cross-domain session loss: Cookies are blocked between different Render subdomains. Solved by switching to localStorage + Authorization Bearer token headers.
  • Render cold starts: Free tier servers sleep after 15 minutes of inactivity. Solved with a keep-alive ping every 10 minutes, retry logic on timeout, and a "server waking up" banner for first-time visitors.
  • AI signal quality: Raw Gemini responses needed structured parsing, confidence clamping (60–85%), and graceful fallback to rule-based signals to avoid empty or nonsensical outputs.
  • SPA routing on Render: Direct URL navigation (e.g. password reset links) returned 404. Solved by configuring Render's Redirect/Rewrite rules to serve index.html for all routes.
  • News API limitations: Free tier APIs either required paid keys or had broken endpoints. Built a CoinGecko trending fallback that generates news articles from live trending coins.

Accomplishments that we're proud of

  • Shipped a fully working AI signal engine backed by Gemini Flash with explainability — not just a confidence score, but why it made the decision
  • Built end-to-end password reset with branded email — fully functional, not a placeholder
  • Achieved 7-day session persistence across page refreshes on a cross-domain free-tier deployment
  • Deployed a production-grade app with real users able to sign up, trade, and use every feature on the live site

What we learned

  • Cross-domain authentication is harder than it looks — cookies have strict rules, Bearer tokens are the pragmatic fix
  • AI outputs need guardrails — clamping, parsing, and fallback logic are non-negotiable for reliability
  • Free-tier cloud infrastructure has real constraints that force creative engineering solutions
  • A layered backend architecture (controllers → services → repositories) pays off immediately when debugging production issues

What's next for TradeX

  • Real brokerage integration (Alpaca, Zerodha) for actual trade execution
  • WebSocket-based live price streaming instead of polling
  • Mobile app (React Native)
  • Portfolio optimization using Modern Portfolio Theory
  • Alerts via SMS/push notifications when signals trigger
  • Multi-user watchlists and social trading features

Built With

  • alpha-vantage-api
  • axios
  • bcrypt
  • coingecko-api
  • css-modules
  • express.js
  • framer-motion
  • google-gemini-flash
  • jwt
  • node.js
  • postgresql
  • react
  • recharts
  • render
  • resend
  • speakeasy
  • supabase
  • vite
Share this project:

Updates