Assaduzzaman Munna

Bachelor of Science in Computer Science and Engineering

North South University (ECE Dept.)

Maintaining a strong academic record with a CGPA of 3.71/4.00 while actively engaging as a Research Assistant and Undergraduate Teaching Assistant.

AI Engineer Intern · The Data Island

Establishing data pipelines and integrating enterprise AI frameworks. Collaborating with the core engineering team to design, test, and optimize scalable machine learning solutions.

• Data Pipelines
• Enterprise AI
• ML Optimization

Undergraduate Teaching Assistant · North South University

Facilitating technical sessions and providing mentorship to undergraduate students in the ECE Department. Supporting faculty in curriculum delivery and grading technical assignments.

• Mentorship
• Curriculum Delivery
• Engineering Principles

On-the-Job Training (AI) · Nippon AI Dojo

Selected participant in a rigorous AI engineering program led by Chowa Giken & AI Samurai Japan. Developed practical skills in AI implementation and model optimization through hands-on group projects.

• Model Optimization
• Practical AI
• Team Collaboration

MIST-ER: Micro-emotion Selective Temporal Emotion Recognition

Developing a lightweight multimodal pipeline (Audio/Video/Text) for micro-emotion classification, achieving 54% accuracy on the MESC dataset. Engineered a cross-modal attention mechanism and optimized the architecture for edge device deployment.

• Multimodal Pipeline
• Cross-modal Attention
• Edge Deployment

Distilled Hybrid Student Framework

Engineered a high-efficiency architecture utilizing Knowledge Distillation, drastically reducing model parameters by 18.5x (1.54M vs 28.6M) for edge deployment while maintaining an exceptional 99.80% accuracy.

• Knowledge Distillation
• MLOps
• Low-Resource Inference

Industrial Automation System (Foil Stamping)

Engineered a deployment-ready automation tool for industrial clients to monitor manufacturing lines, building backend logic with Python and OpenCV to handle high-throughput inspection streams with minimal latency.

• Python
• OpenCV
• System Architecture

Bone Fracture Detection Using Vision Transformers

Accepted at 2nd IEEE Conference on Secure and Trustworthy CyberInfrastructure

Evaluated Vision Transformer architectures (PiT and CaFormer) against traditional CNNs using 4,083 X-ray images. Highlighted the PiT model's superior generalization, achieving 97.51% testing accuracy in automating fracture diagnosis.

• Vision Transformers
• Medical Imaging
• IEEE Publication

Contact Information

I am always open to discussing new engineering opportunities, AI research collaborations, or simply connecting with fellow developers. Feel free to reach out directly.