About the Project

A premium AI platform for verifying news.

Veritas.ai combines classical NLP, machine learning and modern UX to give anyone a fast, explainable verdict on the credibility of online articles.

Dataset

Fake and Real News Dataset — over 40K labeled articles spanning politics, world news and entertainment.

Expected Accuracy

92%+ on held-out test sets, with ensemble models reaching 96% on benchmark data.

Machine Learning Pipeline

01

Preprocessing

Lowercasing, stopword removal, tokenization, text cleaning.

02

Feature Engineering

TF-IDF vectorization with n-gram analysis.

03

Models

Naive Bayes and Random Forest ensembles.

04

Evaluation

Cross-validation with precision, recall and F1 metrics.

Tech Stack

FrontendReact, TypeScript, Tailwind CSS
BackendPython, FastAPI
Machine LearningScikit-learn, Pandas, NLTK
DatabaseMongoDB

API Endpoints

  • POST /predict
  • POST /analyze-url
  • GET /history
  • GET /analytics
  • GET /news-feed
{
  "prediction": "Fake News",
  "confidence": 96.4,
  "explanation": "Contains misleading claims and suspicious language."
}

Developer

Maithreyan D

Designed and engineered end-to-end — from ML pipeline to premium UI.