Projects
Research systems and development projects, from published ML models to shipped web apps.
ACB-TriNet: Malware Classification with Asymmetric Convolutions & Triplet Attention
Dual-branch deep learning system converting malware binaries into three-channel image representations (grayscale, entropy, Sobel edges) for accurate family classification. Achieves 98.98% accuracy on Malimg dataset.
Network Intrusion Detection with Explainable AI
Deep learning architectures for detecting network intrusions, augmented with XAI methods (SHAP, LIME) to explain model decisions, addressing the interpretability gap in network security systems.
Suicidal Ideation Detection Using Language Models
Transformer-based NLP system detecting suicidal ideation in Reddit posts by combining pre-trained language models (BERT, RoBERTa, DistilBERT) with Bidirectional GRU. BERT-BiGRU achieves 95.8% accuracy with only 4.17% false negative rate.
Retinal Disease Classification using OCT Images
Hybrid CNN architectures combining CBAM attention and skip connections with pre-trained backbones (DenseNet121, ResNet50, VGG16, Xception, EfficientNet) for multi-class retinal disease classification. DenseNet-CBAM-Skip achieves 96.28% accuracy.
Borderless Table Detection from Images
Deep learning pipeline using Table Transformer (TATR) to detect and extract tables without visible borders from document images, solving a key gap in document intelligence systems.
Doctor Bhai: Healthcare Assistance Platform
Full-stack web platform connecting patients with doctors for appointment booking and telemedicine consultations in Rajshahi, Bangladesh. Built as a university capstone project.
More experiments, notebooks, and side projects on GitHub
View GitHub Profile