Majid Hosseini, PhD
Artificial Intelligence Research Scientist
Expertise
Multimodal learning, human-centric intelligent systems (multimodal learning, real-world implementation science, federated learning, telehealth.
Biography
Majid Hosseini, Ph.D., is an AI research scientist specializing in multimodal learning, healthcare AI, computer vision, edge AI, privacy-aware sensing, cyber-physical systems, and the Internet of Things.
He earned his Ph.D. in Computer Science from the University of Louisiana at Lafayette in 2025, where his dissertation focused on robust stress detection using real-world data and multimodal learning.
At the University’s Center for Applied AI, Dr. Hosseini develops intelligent systems for patient monitoring, clinical decision support, stress detection, indoor localization, and medical-image analysis. His work has produced real-time clinical vision systems, multimodal health datasets, and privacy-conscious AI solutions developed in collaboration with healthcare and research partners.
He also supervises graduate researchers, co-authors peer-reviewed publications, teaches machine learning and data analytics, and contributes to research proposals.
His research has appeared in venues including Scientific Data, International Forum of Allergy & Rhinology, SN Computer Science, and IEEE and AAAI proceedings. Drawing on experience spanning academia, healthcare, energy, and engineering, he is committed to developing trustworthy AI systems that address consequential real-world problems.
Selected Publications
Machine Learning-Enhanced Clinical Decision Support for Sinusitis Assessment
Focus: Developing multi-class computer vision and rule-based diagnostic frameworks using nasal endoscopy images to standardise sinusitis detection and lower inter-operator variability.
Citation: Gyawali, D., Hosseini, M., et al. (2025). Machine learning-enhanced clinical decision support for diagnosing sinusitis with nasal endoscopy. PMC / Journal Source.
Multimodal Driver Drowsiness & Stress Detection Datasets
Focus: Collecting and processing multimodal data—combining 3D depth facial video, infrared footage, and physiological biometric signals (heart rate, electrodermal activity, blood oxygen saturation)—to monitor physiological states such as driver fatigue and nurse stress in continuous real-world environments.
Citation: Hosseini, M., Sohrab, F., Gottumukkala, R., Bhupatiraju, R. T., Katragadda, S., Raitoharju, J., Iosifidis, A., & Gabbouj, M. (2025). A multimodal stress detection dataset with facial expressions and physiological signals. arXiv. https://doi.org/10.48550/arxiv.2209.13542
Inception-Based Convolutional Recurrent Neural Networks for Video Frame Prediction
Focus: Proposing deep learning architectures (such as Inception-LSTM) that combine variable kernel sizes in gated recurrent units for unsupervised video frame prediction in computer vision systems.
Citation: Hosseini, M., Maida, A. S., Hosseini, M., & Gottumukkala, R. (2020). Inception LSTM for next-frame video prediction (student abstract). Proceedings of the AAAI Conference on Artificial Intelligence, 34(10), 13809–13810. https://doi.org/10.1609/aaai.v34i10.7176
Entropic Complexity Features for Situational Awareness
Focus: Utilizing entropic complexity metrics as features to evaluate non-linear, interconnected behavioral dynamics and improve situational awareness in dynamical systems.
Citation: Ferdowsi, F., Hosseini, M., & Hosseini, M. (2020). Classification of emergent dynamics in microgrids utilizing an entropic complexity-based feature extractor. Data-Enabled Analytics Journal / Conference Proceedings.