About
I'm an R&D Software Engineer at MedAi Limited, where I design AI-driven healthcare solutions focused on human-centered computing. My work is rooted in building culturally-aware, linguistically-inclusive digital health platforms to improve care accessibility for underserved communities.
I’m passionate about bridging the gap between cutting-edge AI research and real-world challenges—particularly in mental health. My interests span machine learning, natural language processing, and user experience design, with a strong emphasis on ethical, inclusive, and human-responsive technologies.
I earned my Bachelor’s degree in Computer Science and Engineering from Bangladesh University of Engineering and Technology (BUET). For my thesis, I worked under the supervision of Professor Dr. Md. Monirul Islam. on Bengali Text Recognition Using Deep Learning, where I built a dataset from printed Bengali documents and trained deep learning models using CNNs, RNNs, and LSTMs, achieving an accuracy of 87.3%.
Currently, my research explores the intersection of generative AI, mental health, and therapeutic visualization—with a focus on how emotionally personalized media can support trauma-informed care.
Research Interests: AI/ML in Healthcare, Multimodal LLM, Mental Health, Human-Computer Interaction, Social Computing, Accessibility in Technology
Professional Experience
R&D Software Engineer | MedAi Health
- Designed and developed AmarDoctor, the first AI-powered digital health platform designed for Bengali speakers
- Developed the architecture for scalable symptom assessment and decision support systems serving thousands of users.
- Manage technical team and drive API design and cloud infrastructure development
- Focus on accessibility and inclusivity in healthcare technology
Teaching | Bangladesh Institute of Science & Technology
- Taught and mentored undergraduate students in Microprocessor and Assembly Languages and Computer Graphics Lab, focusing on hands-on instruction and foundational computing concepts.
- Conducted lab sessions for Microprocessor & Assembly Languages and Network and Information Security, guiding students through practical implementation of low-level programming and secure communication protocols.
- Facilitated interactive, application-driven labs in Computer Graphics and Security, supporting students in applying theoretical knowledge to real-world technical challenges.
Research & Publications
Current Research Focus
Human-Centered AI in Healthcare: My current work centers on developing culturally-aware, linguistically inclusive AI systems for underserved communities, with a particular focus on Bengali-speaking populations. I'm also exploring how generative AI can support mental health therapy—especially through emotionally personalized therapeutic visualizations for trauma-informed care and individuals experiencing emotional regulation challenges.
You can find my concept paper on this topic in the following section. If this research interests you, please send me an email.
Preprints / Manuscript
1. Automatic Speech Recognition of Biomedical Data in Bengali Language
Shariar Kabir; Nazmun Nahar; Mamunur Rashid; Shyamasree Saha (2024). Automatic Speech Recognition for Biomedical Data in Bengali Language. arXiv preprint arXiv:2406.12931. PDF
2. First Multilingual Digital Platform For AI-Driven Primary Care Triage And Patient Management System For Bengali Speakers.
Nazmun Nahar; Shariar Kabir; Sumaiya Tasnia Khan; Suparna Das; Shyamasree Saha; Mamunur Rashid. PDF
3. Through the Mirror of Memory, Healing in Motion: A Generative AI Framework for Trauma-Informed Therapeutic Visualization.
This paper proposes a generative AI framework for trauma-informed therapy, focusing on personalized video synthesis to support inner child healing. By combining therapist-guided prompts with emotion-aware generation, the system aims to create emotionally resonant content that complements therapeutic work—especially for individuals with emotional numbing or limited imaginative capacity. PDF
Key Projects
AmarDoctor - AI Health Platform
Problem: Lack of AI-driven primary care access for Bengali speakers, creating healthcare accessibility barriers for South Asian populations.
Solution: Designed comprehensive medical knowledge graph comprising symptoms, diseases, and weighted relationships; developed clinical decision support algorithms for multilingual symptom assessment and diagnostic recommendations.
Impact: A first-of-its-kind AI-powered platform that has served over 10,000 Bengali-speaking users globally, delivering more than 2,300 personalized healthcare consultations as of now.
Award: AmarDoctor was selected as a Solver Team in the 2024 MIT Solve Global Health Equity Challenge for its AI-powered, culturally inclusive telemedicine platform serving Bengali-speaking communities. The award recognizes its impact in advancing equitable access to primary care.
Symptom Checker and Clinical Descision Support System
Problem: Many patients face difficulty articulating their symptoms and identifying the right specialist.
Solution: Designed an AI-powered symptom checker that suggests relevant symptoms, asks targeted follow-up questions, and provides provisional diagnoses along with specialist recommendations.
Impact: SAchieved 87% diagnostic accuracy validated against physician assessments using 185 simulated patient cases, significantly improving triage efficiency and accessibility to care.
Personalized Food Recommendation System for Disease-Based Dietary Guidance.
Problem: Gap between visual food recognition technology and personalized medical dietary counseling for chronic disease management.
Solution: Integrated computer vision-based food classification with medical knowledge systems for condition-specific dietary filtering and recommendation generation.
Impact: Enables real-time e dietary guidance automation, potentially improving patient adherence to medical dietary restrictions.
Medical Assistant Chatbot Using Rasa
Problem: Limited accessibility to preliminary health assessment tools through natural conversational interfaces, particularly for mental vs. physical health triage.
Solution: Developed intent-based conversational AI using RASA framework with dynamic dialogue management, implementing looped questioning system for symptom classification.
Impact: Demonstrated feasibility of automated health triage through conversational interfaces, enabling preliminary classification between mental and physical health concerns.
Bengali Text Recognition Using Deep Learning
Problem: Lack of effective optical character recognition systems for Bengali printed text, hindering digitization of Bengali literature and documents.
Solution: Developed novel Bengali word image dataset from printed documents, designed and trained deep neural network architectures for character sequence recognition.
Impact: Achieved 87.3% word recognition accuracy, contributing to low-resource language NLP research and digital preservation of Bengali texts.
SaaS Client Management System
Problem: Managing scalable, secure API access for healthcare service providers.
Solution: Built a JWT-authenticated client management platform with customizable service packages and time-based access control.
Impact: Powers scalable B2B healthcare API delivery with robust access and subscription management.