Constructed a cross-platform discovery suite featuring a React/TypeScript web app and a Kotlin/Jetpack Compose Android app, integrating Ticketmaster, Spotify, and GCP Geocoding APIs. Developed robust state management (React Query/State Flow) and persistent storage (MongoDB/Data Store) to sync user favorites, achieving a 90% satisfaction rate across 20+ peer evaluations.

Spearheaded a full-stack gamified progression system to train users in prompt engineering, utilizing Generative AI (Gemini) to evaluate user inputs against a custom quality-scoring algorithm. Launched an automated feedback loop that detects vague inputs and optimizes context windows, successfully gamifying the prompt optimization process for over 10 levels of increasing difficulty.

Led the design of a TensorFlow/Keras CNN-LSTM for real-time ISL recognition, leveraging MediaPipe and OpenCV for hand tracking and achieving 92.8% accuracy (published by Springer). Architected a full-stack ISL platform with a Flask backend serving the AI model and a JavaScript frontend handling real-time video translation across 10+ interactive modules.This project was recognized with the 'Best Project' award for its innovation and impact.

Engineered and optimized a Naive Bayes classifier for a voice-command email system, improving model accuracy by 25% and cutting training time by 30% through data analysis with Pandas and Scikit-learn (published by IEEE). Developed a full-stack voice-to-email application, architecting a Django backend to integrate the ML model and utilize Pythons SMTP library for sending emails, and a JavaScript frontend for voice command capture and audio feedback via gTTS.

Spearheaded a number plate OCR pipeline, boosting Pytesseracts accuracy to 82% on 1,000+ vehicles with a custom Keras model, earning recognition at the Smart India Hackathon (publication by IEEE). Launched a full-stack recognition application, creating a PHP backend on an Apache server to expose the Python computer vision system via an API and to visualize model performance graphs generated with Matplotlib.

Architected and launched a robust eCommerce website from the ground up, utilizing React for a dynamic front-end and MySQL with XAMPP for the back-end. I developed essential functionalities, including secure user sign-up and login, a home page, and individual product pages. By focusing on a clean and user-friendly UI/UX design, I achieved a 30% boost in user engagement.

Spearheaded the development of a real-time age and gender identification system, achieving 85% classification accuracy. I trained a Convolutional Neural Network (CNN) using Keras, Dlib, and OpenCV on a classified Kaggle dataset, enabling the final program to deliver dynamic and targeted content based on live facial analysis.

Created a comprehensive, full-stack portfolio management platform by developing a robust backend with Python and SQLite, and a responsive frontend using HTML, CSS, and JavaScript. The applications core feature is an integrated predictive analytics module that leverages linear regression algorithms to forecast future asset performance based on historical data. This system provides users with actionable insights through real-time data analysis and dynamic visualizations, empowering strategic investment planning with a validated trend prediction accuracy rate of 90%.

Designed and launched a full-stack restaurant reservation platform, developing a robust backend with Java and SQL and a highly intuitive, easy-to-navigate frontend using HTML, CSS, and JavaScript. The project emphasized a clean and user-centric design, creating a seamless workflow that simplified the booking process for customers. This commitment to a superior user experience directly resulted in a 40% reduction in the average time required to secure a reservation.