Project 1: AI Sudoku Solver & Generator
A full-stack AI-powered Sudoku Web App that allows users to generate, solve, and validate puzzles of varying grid sizes (4x4, 9x9, 16x16). The app features both a Constraint Satisfaction Problem (CSP) solver for all grid sizes and a Neural Network solver specifically for 9x9 puzzles. Players can track their scores, upload avatars, and view a global leaderboard. Built using Flask (backend), HTML/CSS/JS (frontend), and SQLite (for storing scores).
Interactive interface of the AI Sudoku web app with generation, solving, and leaderboard features.
Multi-size Support
Play and solve 4x4, 9x9, or 16x16 Sudoku puzzles with adjustable difficulty.
AI Solvers
Backed by CSP for logical solving and a neural network model trained specifically for 9x9 puzzles.
Real-time Validation
Validate puzzle progress live and get immediate feedback on correctness.
Leaderboard
Track and display top scores with time, difficulty, and optional avatars.
Avatar Upload
Upload custom avatars to personalize your leaderboard appearance.
SQLite Integration
Store player performance securely using SQLAlchemy and SQLite.
Project Demonstration
Watch how the AI Sudoku Solver works with CSP and Neural Network methods in real-time.
Project 2: Plant Disease Detection
This project uses convolutional neural networks to classify plant diseases across 14 crop types and 39 different disease classes with remarkable accuracy. It serves farmers and agricultural researchers for early disease detection using image-based diagnosis.
14 Crops, 39 Classes
Supports Tomato, Apple, Corn, Potato, Grape, and more. Distinguishes both healthy and diseased states.
Trained Deep Models
Uses CNN, ResNet18, MobileNetV2, and EfficientNetB0 architectures. Models trained on 54K images using PyTorch.
99% Accuracy
Tested and validated with early stopping, checkpointing, and visualization of training vs. validation performance.
Flask Web App
Complete user interface with multi-model selection, history tracking, real-time predictions, and supplement advice.
CSV Integration
Displays predicted class, disease description, prevention steps, and supplement information fetched from CSVs.
Real Image Support
Works with camera or gallery images. Robust predictions after training on real-world leaf image dataset.
Screenshots
User uploads a leaf image .
Selects model.
Receives disease name, description, and prevention steps.
Confusion matrix comparison of CNN. Validated with high accuracy.
Confusion matrix comparison of ResNet18. Validated with high accuracy.
Confusion matrix comparison of MobileNetV2. Validated with high accuracy.
Confusion matrix comparison of EfficientNetB0. Validated with high accuracy.
Project Demonstration
<Watch the video to see how the Plant Disease Detection system works in real-time.
Project 3: Personal Portfolio Website
A responsive and elegant multi-page portfolio crafted using HTML, CSS, and JavaScript. This website reflects my design philosophy—clean, user-centric, and approachable. It showcases my technical projects, resume, experience, and contact information while ensuring mobile-first responsiveness and light/dark theming.
Elegant UI
Soft colors, rounded sections, and gentle animations give the site a welcoming feel.
Clean Codebase
Structured HTML, modular CSS, and reusable design patterns ensure maintainability.
Fully Responsive
Optimized for desktops, tablets, and mobile devices with fluid layouts.
Dark Mode Toggle
Built-in toggle to switch between light and dark themes for user preference.
Resume Download
One-click access to my resume via a well-placed CTA on the resume page.
Personal Touch
Branded with my name, personality, and visuals to reflect who I am as a developer.
Screenshot of my personal portfolio website—designed for clarity and engagement.
Project Demonstration
A short walkthrough of my personal portfolio site, layout, and responsive features.