Hasan Jehangir

Real-time tonic-clonic seizure detection system using multimodal biomedical signals (EMG, PPG, IMU). Implements ChronoNet architecture for temporal pattern recognition and deployed on edge devices for clinical monitoring.

Achieved 94% sensitivity in seizure detection with minimal false positives. Uses signal preprocessing, feature extraction, and neural classification.

CNN-based system for automated lung disease classification using respiratory sound analysis. Employs MFCC feature extraction and attention mechanisms for improved diagnostic accuracy.

Classifies conditions including pneumonia, asthma, and COPD. Integrated into healthcare applications for non-invasive screening.

Deep learning classifier for melanoma detection from dermoscopic images using transfer learning and ensemble methods for clinical reliability.

Combines ResNet and VGG architectures with voting ensemble. Designed for dermatology clinics with interpretable predictions.

Intelligent question-answering system providing authentic responses to Islamic queries using the Quran and verified Islamic books. Employs RAG architecture with semantic search and LLMs.

Deployed production system handling thousands of queries. Retrieves from verified sources and generates contextually accurate answers in multiple languages.

Computer vision system for automated recyclable item classification using YOLO object detection with real-time processing capabilities.

Detects 20+ recyclable material types. Integrated with IoT sensors for smart waste management applications.

Advanced computer vision project using attention-based encoder-decoder networks for restoration and enhancement of historical artifact images.

Combines domain-specific loss functions with neural attention mechanisms. Preserves fine details while removing degradation.