🎓 PhD in Informatics (Shizuoka University, 2024)
🔬 Research Focus: NLP • Multimodal ML • Vision-Language Models • Explainable AI
📍 Currently: Product Development Engineer at Pi Photonics, Japan
-
Cross-Lingual Vision-Language Models: Evaluating CLIP's performance on Japanese vs English memes (repo)
-
Dual-Rationale Attention Mechanisms: Novel architecture for sarcasm-aware hate speech detection
-
Industrial AI: OAK-D camera-based safety monitoring systems
Currently developing AI systems for industrial applications:
- Edge AI for Safety Monitoring: Camera-based crane detection/ wave detection/ human detection on embedded platforms
- Real-time Video Analysis: Feature extraction and object detection on STM32 microcontrollers
- Firmware Development: Implementing ML inference pipelines on resource-constrained devices
Combining research expertise in deep learning with hands-on embedded systems deployment.
-
Mamun, M.B., Tsunakawa, T., Nishida, M., Nishimura, M. (2024). "Hate Speech Detection by Using Rationales for Judging Sarcasm." Applied Sciences, 14(11), 4898.
-
Mamun, M.B. (2025). "Sarcasm-Aware Hate Speech Detection Using Rationales." Journal of the Japanese Society for Artificial Intelligence, 40(1), 1058-1059.
Deep Learning & NLP: PyTorch • Transformers (BERT, RoBERTa, DeBERTa, CLIP) • Attention Mechanisms • Multi-task Learning • Vision-Language Pre-training
Computer Vision & Edge AI: OpenCV • OAK-D Depth Cameras • Real-time Object Detection • Video Processing • Embedded Inference
Embedded Systems: STM32 • Raspberry Pi • Firmware Development • Real-time Feature Extraction • Industrial Safety Applications
Tools & Frameworks: Hugging Face Transformers • scikit-learn • TensorFlow • C++ • Python
💡 Note: My current industry work at Pi Photonics is maintained on the company's private GitLab server. This GitHub showcases my independent research in multimodal ML and NLP.