ai★ Featured CreationDistilBERTPyTorchNext.js 15FastAPINLP
VerifiNews-AI – Fake News Detection System
Project Brief & Architecture
VerifiNews-AI is an AI-powered news verification platform that helps users identify misinformation by analyzing news articles in real time. Built with Next.js, FastAPI, and Supabase, it leverages DistilBERT and MobileBERT for fake news detection, VADER for sentiment analysis, and integrates live news from the GNews API along with real-time cricket scores via the Cricbuzz API. The platform features secure authentication, analysis history, and a modern, responsive interface for an intuitive user experience.
Technical Case Study
VerifiNews-AI Technical Case Study
Executive Summary
VerifiNews-AI is a full-stack deep learning solution engineered to detect misleading articles, manipulated claims, and fabricated news stories in real time.
Core Architecture & Deep Learning Stack
- Transformer Backbone: Fine-tuned DistilBERT (Hugging Face Transformers) achieving 82%+ validation accuracy across 10,000+ labeled articles.
- Preprocessing Pipeline: Custom text cleaning pipeline featuring Tokenization, Lemmatization, Stopword Removal, and TF-IDF n-gram vectorization.
- Backend API: Python FastAPI microservice with asynchronous batch inference and Redis caching.
- Frontend Dashboard: Next.js 15 App Router with live credibility score gauges, named-entity recognition (NER) highlights, and sentiment heatmaps.
Key Engineering Features
- 1Real-time Claim Verification: Sub-500ms latency inference on input URLs and raw articles.
- 2Explainable AI Highlights: Visual token attention weights highlight suspect phrases.
- 3Automated Source Indexing: Cross-references claims against trusted knowledge graphs.
Project Connections
Lighthouse Audits
Performance98%
SEO Optimization96%
Security Audit97%
Specifications
Categoryai
StatusCompleted
Tech Stack
DistilBERTPyTorchNext.js 15FastAPINLP