Selected research and engineering projects.
Lightweight fetal ultrasound segmentation with DLCE hybrid loss, teacher–student distillation, and GradCAM++ explainability.
Hybrid MFCC, log-mel, chroma, and cochleagram features with CNN-Transformer and CNN-GRU models for Bangla emotion recognition.
Adapted ehrMGAN for pediatric sepsis and built structured patient-hour level clinical data pipelines.
Time-series forecasting using LSTM and GRU models for stock movement prediction.
Machine learning models to identify students at risk and analyze performance factors.
Full-stack book sharing platform using Flask, SQLAlchemy, and MVC-based architecture.