Mohua Akter
CSE graduate specializing in Data Science and Machine Learning,
East West University
Project
Designed the RAG pipeline and hybrid RAG+LLM chatbot, moving from Gemma 3 to Qwen; evaluated five Rasa, RAG+LLM, and hybrid architectures for Bangla, English, and Banglish in low-resource settings. Best architecture achieved the strongest semantic-similar
Project
Built a RAG admission chatbot and fine-tuned Gemma 3 on custom JSONL data covering EWU programs, tuition, policies, and the admissions calendar.
Project
Built a web attendance system with real-time face recognition and role-based student, teacher, and administrator portals.
Project
Benchmarked DeepLabV3, SegFormer, and YOLOv26 on 1,969 images across 30 fruit classes; YOLOv26 achieved 0.944 mIoU, outperforming both alternatives. Published processed train/validation/test splits as public Kaggle datasets.