VeritasAI — Rumour Verification System
Engineered and deployed a real-time rumour verification system achieving 99.43% accuracy using a Passive Aggressive Classifier with TF-IDF vectorization trained on 44,000+ labelled news articles. Integrated NewsAPI for live evidence retrieval and shipped a Streamlit app with claim checking, article analysis, a filterable live news feed, and URL-based article extraction. Diagnosed and corrected dataset bias and added confidence scoring to surface model uncertainty.
Document Intelligence Analyzer
Architected a GenAI-powered document intelligence system combining OCR and LLMs for automated analysis, fraud detection, and conversational document querying. Designed an OCR extraction and text-preprocessing pipeline and deployed the application with JSON export on Streamlit Cloud.
Smart Crop Health & Soil Detection System
Trained a CNN image classifier to distinguish healthy and diseased crops using data augmentation and hyperparameter tuning. Developed a soil-parameter analysis module and deployed a Flask web app serving real-time crop-health predictions and soil insights.
Movie Recommender System
Engineered a content-based recommendation engine using TF-IDF vectorization and cosine similarity. Processed movie genres, cast, crew, and keywords and optimized similarity computation for fast recommendations across a large movie dataset.