Designed and developed a hierarchical waste classification system using a Multi-Branch CNN architecture with a shared feature extraction backbone. Performed multi-level classification into organic/inorganic waste, material categories, and specific waste sub-types while reducing redundant computation. Improved computational efficiency by sharing convolutional layers across multiple prediction branches.

Role-Based Survey Web Application— Developed a responsive survey platform using React, Firebase Authentication, Firestore, and Tailwind CSS with separate Admin and User roles. Admins can create surveys, manage questions, and view response analytics, while users can access assigned surveys and submit responses securely.