Research Projects
EfficientU-Net for Breast Tumor Analysis
A novel deep learning method for breast tumor segmentation in ultrasound images, combining EfficientNetB7 as an encoder with an Atrous Convolution block to improve segmentation accuracy.
Technologies
Highlights
- Developed a lightweight architecture combining EfficientNetB7 with U-Net
- Implemented Atrous Convolution blocks for better feature extraction
- Achieved state-of-the-art performance on breast ultrasound datasets
- Published in Neural Processing Letters (2023)
Associated Papers
Links
UMA-Net for Breast Ultrasound Segmentation
Advanced deep learning architecture incorporating multi-scale attention mechanisms and adaptive ensemble loss for accurate breast tumor segmentation in ultrasound images.
Technologies
Highlights
- Proposed novel multi-scale attention mechanisms for better feature extraction
- Developed adaptive ensemble loss function for improved training stability
- Achieved superior performance on benchmark datasets
- Published in Medical & Biological Engineering & Computing (2025)
Associated Papers
Links
Genetic Algorithm for Feature Selection
An ensemble approach combining deep learning and genetic algorithms for optimal feature selection in breast ultrasound image classification.
Technologies
Highlights
- Developed novel feature selection method using genetic algorithms
- Integrated with deep learning models for improved classification
- Achieved 4-9% improvement in accuracy on benchmark datasets
- Published in Image and Vision Computing (2024)
Associated Papers
Links
Shifa.AI - AI-Powered Healthcare Platform
An advanced healthcare platform providing AI-powered medical assistance, symptom analysis, and personalized health assessments with secure data management.
Technologies
Highlights
- AI-powered symptom analysis and health assessments using Google Gemini AI and OpenAI
- Secure user authentication with JWT and NextAuth.js integration
- Responsive UI built with Tailwind CSS and Headless UI components
- MongoDB database with Mongoose for secure medical data management
- Modern development setup with TypeScript and Next.js API routes
- Real-time toast notifications for user feedback
UGC NET CS HUB
A specialized e-learning platform tailored for aspirants preparing for the UGC NET (University Grants Commission - National Eligibility Test) in Computer Science and Applications.
Technologies
Highlights
- Comprehensive question bank with detailed solutions and explanations
- Practice modes including subject-wise, topic-specific, and difficulty-based questions
- Performance analytics and progress tracking for effective learning
- Secure authentication with role-based access control
- Admin dashboard for content management and user monitoring
GoalTrackr - Personal Goal Management
A full-stack web application for tracking personal goals, tasks, and progress with visualization dashboards and journaling features.
Technologies
Highlights
- Built with modern web technologies including Next.js and MongoDB
- Implements secure user authentication and data persistence
- Features interactive dashboards with progress visualization
- Live demo available at smart-goal-tracker.vercel.app
IEEE BigMM Data Challenge
Multi-task multimodal framework for predicting labels from tweets, developed for the IEEE BigMM Data Challenge.
Technologies
Highlights
- Developed solution for multi-label tweet classification
- Incorporated both text and image modalities
- Implemented state-of-the-art transformer architectures
- Competed in IEEE BigMM conference challenge