Built a student stress prediction system using SVM, Random Forest, and XGBoost achieving ~85% accuracy through optimized feature engineering and hyperparameter tuning.

• Implemented and compared SVM, Random Forest, and XGBoost classifiers
• Feature engineering and hyperparameter tuning for optimal accuracy
• Achieved ~85% prediction accuracy on real student data

Actively contributing to collaborative creative initiatives and technical event participation within VIT's film community.

• Collaborative creative project development
• Technical event coordination and participation
• Cross-team communication and creative ideation

IoT-powered intelligent attendance system combining ESP32-CAM real-time face capture, deep learning face recognition, anti-spoofing detection, and automated attendance logging with LED/buzzer feedback.

• Real-time face detection and recognition via ESP32-CAM hardware
• Anti-spoofing CNN model detecting live vs photo/screen attempts
• Automated Excel attendance logging with timestamp

Enterprise-grade online voting platform with role-based access, JWT authentication, one-vote enforcement, and secure MongoDB-backed election management.

• Role-Based Access Control (RBAC) with admin/voter separation
• JWT-authenticated secure sessions with duplicate vote prevention
• Real-time election result tracking and management

Client-side intelligent interface system that adapts in real-time to user interaction patterns without any backend or data storage.

• Real-time behavioral pattern analysis (typing, mouse, corrections)
• Dynamic UI state adaptation across four behavioral modes
• Zero backend — fully client-side with complete privacy

Machine learning system predicting student stress levels from lifestyle and academic factors using optimized feature engineering achieving ~85% accuracy.

• Multi-model comparison: Logistic Regression, RF, SVM+PCA, XGBoost
• Feature engineering and hyperparameter tuning across models
• XGBoost achieved ~85% prediction accuracy