The problem

Traders face information overload analyzing disparate stock feeds, financial news articles, and trading journal entries across separate tools.

The approach

Built a Next.js application integrating WebSocket streaming from Finnhub and Yahoo Finance with Hugging Face NLP models to automate market sentiment scoring and P&L analytics.

Architecture

Frontend

Next.js, TypeScript, React 19, Recharts financial graphics, Zustand state container.

Backend

Node.js REST service with WebSocket listeners fetching market quote tickers.

Database

MongoDB storing historical trading logs, sentiment analysis caches, and watchlists.

Authentication

JWT authentication with HTTP-only refresh tokens.

Key decisions

  • Zustand state caching for high-frequency market tick data to avoid unnecessary React re-renders.
  • Client-side caching layer for Hugging Face NLP sentiment scores to minimize third-party API rate limits.

Working through the challenges

  • Maintaining smooth 60fps chart rendering during high-frequency WebSocket updates.
  • Normalizing disparate REST payloads from multiple market data providers.

What the application supports

  • WebSocket live ticker feeds for equities, market indices, and commodities.
  • Hugging Face NLP news sentiment scoring.
  • Interactive trading journal with automated P&L analytics and win-rate charts.