PhD Candidate in Computer and Information Science | Machine Learning & AI Researcher | Thermal Imaging Specialist

Hello! I'm Kais Riani, a Machine Learning Engineer at Astemo, specializing in machine learning, thermal imaging, and human behavior analysis. My research focuses on AI-driven circadian rhythm modeling, thermal facial landmark detection, and non-contact physiological signal extraction. With experience collaborating with Toyota Research Institute and Ford Motor Company, I design multimodal AI systems for autonomous systems, healthcare, and security applications.

• Enhancing PyraMoT, an optimized thermal facial landmark detection model using MobileNetV2 + FPN.
• Developing AI-driven models for non-contact physiological signal extraction and human behavior recognition.

• Self-supervised learning techniques for thermal imaging applications.
• Generative AI models for data augmentation and human state recognition.
• Cloud-based AI deployment using AWS, Azure, and Google Cloud.

• AI-driven human behavior analysis projects.
• Thermal imaging applications for biometrics, security, and healthcare.
• Deep learning models for non-contact physiological monitoring.

• Exploring federated learning for privacy-preserving behavioral analysis.
• Improving thermal image segmentation with self-supervised and transformer-based architectures.

• Thermal imaging in AI & Machine Learning.
• Human circadian rhythm analysis using deep learning.
• Facial landmark detection & segmentation techniques.

• Email: kriani@umich.edu
• LinkedIn: linkedin.com/in/kais-riani
• GitHub: github.com/RianiKais
• Website: sites.google.com/umich.edu/kais-riani

• Thermal imaging isn't just for night vision! It can detect stress, emotions, and fatigue through micro-temperature variations.
• I’m passionate about AI for human behavior analysis and how non-contact sensing can revolutionize healthcare and security.

• Designed PyraMoT, integrating MobileNetV2 + FPN, reducing thermal landmark detection errors by 10%.
• Developed using TensorFlow and PyTorch to improve thermal face tracking in occluded environments.

• Developed PyraSegNet, a hybrid deep learning model for thermal facial segmentation.
• Achieved 96.30% F1 score and 93.14% IoU in segmenting whole face, forehead, eyes, cheeks, and nose.

• Created T5050-FR, a large-scale annotated thermal dataset with diverse demographics and occlusions.
• Designed for thermal segmentation, biometrics, and behavioral research.

• Developed a multimodal AI system integrating thermal imaging, visual data, and physiological signals.
• Achieved 77% accuracy in circadian rhythm state classification, contributing to autonomous vehicle safety.

• Built a deep learning pipeline to extract 3 physiological signals from thermal imagery.
• Achieved 70% accuracy in enervation state classification, offering an alternative to contact-based sensors.

• Das, K., Papakostas, M., Riani, K., et al. (2022). Detection and Recognition of Driver Distraction Using Multimodal Signals, ACM TiiS.
• Kamboj, M., Hessler, C., Riani, K., et al. (2020). Multimodal Political Deception Detection, IEEE MultiMedia.

• Riani, K., Sharak, S., Abouelenien, M. (2024, May). PyraMoT: A Novel Framework for Enhanced Facial Thermal Landmark Detection, IEEE FG 2024.
• Riani, K., et al. (2023). Non-Contact Based Modeling of Enervation, IEEE FG 2023.
• Lilley, M., Das, K., Riani, K., et al. (2022). A Topological Approach for Facial Region Segmentation in Thermal Images, IEEE ISM 2022.

Programming & AI Frameworks:

Data & Cloud Tools:

Development & Version Control:

🔗 Let’s connect and collaborate on innovative AI & thermal imaging projects! 🚀