My Projects

Lecture Voice-to-Notes Generator

AI-powered Streamlit platform that converts long-form lecture audio into structured notes, quizzes, and flashcards using Whisper + transformer summarization models with GPU-tuned ASR/VAD.

Modular NLP components with caching and parallel chunking deliver 5× faster runs on 30+ minute inputs, while the interface adds progress tracking, Fast Mode, and PDF/DOCX/JSON exports.

AI-Assisted Dermatological DSS

EfficientNet-B0 classifier covering 7 skin lesion classes with 81% validation accuracy and 0.67 macro-F1 on HAM10000, accelerated via AMP and optimized Albumentations pipelines.

Integrates Grad-CAM explainability, macro-F1 driven evaluation, and a <50 ms Streamlit inference UI for clinician-focused prototyping.

AI Interview Assistant

End-to-end automated interview workflow that parses resumes, extracts candidate insights, and generates timed AI-driven questions powered by Groq Llama 3.1.

Recruiter dashboard ships with sortable tables, transcript modals, PDF export, and Redux-backed scoring persistence—no backend required.

UML Designer AI

Converted GitHub repos into interactive UML class diagrams across 7+ languages using AST parsing + AI enrichment; reduced manual diagramming effort by ~80% with one-click export.

AI Network Traffic Shaper

Built an end-to-end pipeline on Windows: traffic generation (VoIP/FTP/HTTP), PyShark/TShark capture to CSV, sklearn model (~0.999 acc), and real-time shaping via Windows Firewall with safe rollback.

Psychometric Test System

An intelligent system to assess cognitive abilities and personality traits using structured psychometric evaluations with data-driven insights.

Travel Planner

Efficient route planning using shortest path algorithms to minimize travel costs and time with intelligent pathfinding.

YOLOv5 Real-Time Object Detection

Colab-based object detection pipeline using YOLOv5 pretrained models, enabling real-time image capture from laptop cameras with on-the-fly object recognition.

College Lost & Found Application

Developed and deployed a full-stack MERN + Cloudinary platform supporting 200+ campus users, cutting lost-item recovery time by 60% with streamlined reporting and secure JWT flows.

Enhanced backend efficiency by 35% through modular Express controllers, indexed MongoDB queries, and Cloudinary-driven media storage designed with RESTful best practices.

Increased user engagement by 45% via a responsive Tailwind UI, reusable components, and real-time status updates that raised report completion and item return confirmations.

AgriSure: Agricultural Insurance Platform

Full-stack agricultural fintech platform using Flask, enabling 500+ farmers to manage crop insurance and loans with role-based access control, reducing manual errors by 70%.

Intelligent Search Platform

Full-stack AI-powered search tool using FastAPI and React, expanding queries with NLP (T5/BART) and integrating Google/YouTube APIs with multithreaded backend for enhanced performance.

Fog4Det: Vehicle Fog Detection System

Real-time fog detection system using FastAPI and Next.js with dual-mode interfaces, implementing OpenCV-based analysis to classify visibility conditions and provide driving recommendations.

Reddit Stock Sentiment Analyzer

Flask-based NLP dashboard analyzing Reddit posts from financial subreddits using VADER sentiment analysis, with interactive dark-mode UI and Matplotlib visualizations.

Fake News Detection Model

A deep learning model trained to classify news articles as fake or real based on linguistic and contextual cues using advanced NLP techniques.

Dining Philosopher & Word Ladder

Implementation of classic computer science problems: Dining Philosopher problem for OS synchronization using semaphores, and Word Ladder problem using graph traversal algorithms with optimized pathfinding strategies.