Jannatul Ferdous Salma

Detail-oriented Assistant Software Engineer with hands-on experience at Millennium Information Solution Ltd., excelling in manual and automation testing. Proficient in API testing and Selenium, I effectively enhance product stability through rigorous defect tracking and collaboration. Strong analytical skills drive my commitment to quality assurance and continuous improvement in software development.

1. Leveraging BanglaBERT: A Transformer-Based Approach for Multiclass Hate Speech Detection in Bangla Social Media.

• Institution: Adi Shankara Institute of Engineering and Technology, Kalady
• This study proposes a BanglaBERT-based deep learning model for detecting hate speech in Bengali, addressing script complexity and data scarcity through advanced preprocessing and tuning. It achieves 93.21% accuracy, outperforming traditional methods, and supports real-time content moderation in Bengali NLP.

2. Automating Dental Diagnosis: A Machine Learning Approach to Efficient and Personalized Detection of Dental Problems X-Rays.

• Institution: International Conference on Data Mining and Information Security (ICDMIS 2024)
• This study develops a custom Convolutional Neural Network (CNN) achieving 94.13% accuracy in identifying dental issues such as cavities, implants, and impacted teeth from X-rays. Outperforming pre-trained models like DenseNet201 (74.9%) and YOLOv9 (81.3%), the approach enhances diagnostic speed and accuracy, reducing workload for dental professionals while ensuring timely, personalized patient care

3. Load Balancing Algorithm: A Markov Chain Model Using Queuing Theory Approach.

• Institution: Innovation in Science, Engineering, and Technology (ICISET 2024)
• This study proposes a custom load balancing algorithm to optimize task scheduling in data centers, outperforming Round Robin by controlling execution time growth. Using Markov Chain simulations, the study analyzed parameters like server utilization and queue length, offering insights to enhance data center efficiency.