- Meetbot — Built a full-stack AI meeting assistant that automates transcription, summarization, task extraction, scheduling, and live meeting participation using Groq Whisper, Gemini, Flask, React, and Firebase.
- SEO Intelligence Tool — Full-stack app using NLP (KeyBERT, spaCy) + OCR (Tesseract) to extract SEO insights from articles and PDFs. Automated keyword extraction and meta generation, reducing manual analysis time by 70%.
- StyleTune: AI Fashion-to-Music Recommender — A Streamlit app using Google Gemini AI to recommend Spotify songs based on clothing images. Processed 5000+ tracks, boosting engagement by 20%.
- Flexion: AI-Powered Fitness Coach — Real-time posture correction using TensorFlow with a React front-end and FastAPI backend. Integrated GPT-2 for adaptive workout plans and reduced latency by 30%.
- Micro-Doppler Target Classification — Deep CNN model for drone-vs-bird radar classification using spectrogram analysis. Built a React dashboard with 5+ performance visualizations.
- Financial News Sentiment Analysis — Scraped live Yahoo Finance news and used FinBERT + PyTorch to classify sentiment. Visualized insights using matplotlib, seaborn, and pywaffle.
- Automated Grading System — BERT-based grading app built with PyTorch + Streamlit. Reduced grading effort by 70% and improved accuracy using K-means clustering + cosine similarity.
- Thyroid Disease Prediction — Achieved 0.98 accuracy using HistGradientBoostingClassifier. Preprocessed medical data, handled missing values, and saved model via Joblib.
- Phishing Email Detector — Machine-learning based phishing detector with NLTK preprocessing and a Streamlit UI. Trained a Multinomial Naive Bayes model with TF-IDF features.
- Tech Stocks Data Analysis — Analyzed stock trends with moving averages, daily returns, and visualizations using matplotlib & seaborn.
- Library Management System — A Python-based system for managing books, users, and transactions.
