PakNews Bias Scanner — AI News Analysis

Full-Stack (TypeScript + React Native)

About

AI-powered news analysis platform that detects political bias in Pakistani news articles using ML models. Multi-platform app with automated RSS feed aggregation, ONNX-based bias classification, and real-time article clustering by political topic.

My Role

Built the full platform: Fastify backend with ONNX runtime for bias inference, React Native Expo mobile app with infinite scroll and bias filtering, web-based labeling tool, and ML training pipeline.

Core Capabilities

• ML-based bias classification (Pro-PTI, Pro-PMLN, Pro-PPP, Pro-Establishment, Neutral)
• RSS feed aggregation from multiple Pakistani news sources
• Real-time article clustering by political topics
• Interactive labeling tool for manual bias annotation
• Scheduled 30-minute feed fetch and topic clustering jobs
• Mobile app with infinite pagination and bias filtering

Architecture

• TypeScript Fastify backend with ONNX Runtime for ML inference
• React Native Expo SDK 55 with Expo Router
• Supabase for database and authentication
• Node-cron for scheduled feed fetching and clustering
• Python pipeline for ONNX model training and export

Tech Stack