Research Engineer in NLP & Computer Vision Advancing Low-Resource Language Models and Medical AI

LinkedIn Scholar ORCID Scopus Email

I’m a Research Engineer specializing in NLP and Computer Vision, developing transformer-based, explainable AI systems for healthcare and low-resource language understanding.

Research Areas:

• AI for Healthcare & Public Health Risk Assessment (SRMH)
• Vision-Language & Multimodal Learning
• Low-Resource Bangla NLP (NER, Hate Speech, Dialect Modeling)
• Explainable AI (XAI) for Medical Imaging
• Generative & Transformer-based Modeling

🤖 Bangla Medical GPT — Adolescent Health Chatbot
An LLM-powered assistant offering context-aware, confidential support for adolescent health and wellbeing.

🧬 AI-based SRMH Risk Assessment Tool
An AI-driven predictive model assessing sexual, reproductive, and mental health risks among populations aged 14–49 in LMIC contexts.

🖼️ VisionLLM for Low-resource Image Captioning
A transformer-based image captioning framework improving accessibility for Bangla-speaking users in low-resource environments.

Systematic review on AI-driven tools for sexual, reproductive, and mental health (SRMH) risk assessment.
Paper
BMC Medical Informatics and Decision Making, Springer Nature (2025)
CITATION.cff included → use GitHub’s Cite button for BibTeX/APA/MLA.

A regional Bangla NER benchmark enhancing domain adaptation across linguistic variations.
Paper
Under Review at PLOS ONE (2025)

A transformer model detecting hazardous selfie behaviors to promote safety awareness.
Paper
IEEE WIECON-ECE 2024, Chennai, India

A multimodal VQA framework integrating CNN–LSTM architecture for Bengali language understanding.
Paper
IEEE ICCIT 2022, Cox’s Bazar, Bangladesh

➡️ Full list → Google Scholar

A large-scale Bengali Visual Question Answering (VQA) dataset advancing multimodal understanding in low-resource languages.
Dataset
CITATION.cff included → use GitHub’s Cite button for BibTeX/APA/MLA.

A regional Bangla NER corpus capturing dialectal and contextual variations across Bangladesh for improved linguistic adaptability.
Dataset

A curated computer vision dataset for identifying extreme vs. normal selfies, supporting studies in digital safety and behavior analysis.
Dataset

Open to collaborations in:

• Low-resource NLP (Bangla NER, Dialects, Hate Speech)
• Healthcare AI & SRMH Risk Assessment
• Explainable Multimodal Learning and Vision-Language Systems

LinkedIn Kaggle ResearchGate Semantic Scholar

Views

© Shifat Islam — Research Engineer in NLP & Multimodal AI. For citation or reuse, please refer to each project’s CITATION.cff or DOI.