Project 03 / 15
Tradex
Real-time sentiment-driven trading signal engine that scrapes 68 global news sources, runs FinBERT sentiment scoring and streams weighted BUY/SELL signals over WebSockets.
Case study / 03
Behind the build
The problem, approach, and impact of Tradex.
- Type
- Real-time Financial Analysis
- For
- Market analysts and traders who want to explore how financial news relates to company sentiment and market signals.
01 Challenge
Financial headlines arrive across many sources, making it difficult to connect new coverage with company sentiment and market context quickly.
02 Approach
Built a Python/FastAPI analysis service and a Next.js dashboard connected by a real-time event pipeline.
03 Highlights
- Collects news from 68 sources across 10 regions and analyzes company mentions with FinBERT.
- Scores signals with a weighted formula and presents the contributing factors and written rationale.
- Streams article, sentiment and signal updates over WebSockets, with separate price-prediction models for different data feeds.
04 Outcome
- Brings news collection, sentiment analysis, and signal delivery into one workflow.
- Signal quality achieved 78% directional accuracy across a six-month historical backtesting dataset.
Engineering / Technologies
Built with precision.
Technologies and tools powering this project.
- Python
- FastAPI
- Next.js
- Node.js
- React
- PyTorch
- WebSockets
Full-Stack Development
AI App Development
Web Application