Short Bio

Mohammad Mahdi Azarbeik is a researcher at the Ludwig Boltzmann Institute Digital Health and Patient Safety (LBI DHPS) and a PhD candidate in Computer Science (Informatics) at TU Wien. His work sits at the intersection of reinforcement learning and critical care medicine, where he develops data‐driven decision‐support tools to optimize treatment strategies for intensive care patients.

He earned a BSc in Mechanical Engineering from K. N. Toosi University of Technology and an MSc in Mechanical Engineering from Sharif University of Technology. For his master’s thesis, “Augmenting inertial motion capture with SLAM using EKF and SRUKF data fusion algorithms,” he combined sensor fusion techniques with simultaneous localization and mapping to enhance motion‐capture accuracy in dynamic environments.

PhD Project - Decision Support in Intensive Care

Supervised by Clemens Heitzinger

Mahdi’s doctoral research aims to advance reinforcement‐learning–based decision support for critically ill patients by:

• Enhancing state representations: Moving beyond clustering to incorporate temporal and multimodal data (e.g., vitals trajectories, lab trends) for richer patient modeling.
• Exploring advanced RL architectures: Investigating actor‐critic and model‐based methods to improve learning stability and sample efficiency in highly stochastic ICU environments.
• Integrating domain knowledge: Embedding clinical guidelines and physiological constraints into reward structures and policy learning to ensure safety and interpretability.
• Bridging retrospective and prospective evaluation: Designing real‐time, adaptive evaluation frameworks that can transition RL policies from offline validation to live clinical decision support, ultimately aiming to improve patient outcomes in sepsis, AKI, and beyond.

Publications and Conferences

Journal Papers

• L. Kapral, M. M. Azarbeik, R. Weiss, R. Bologheanu, C. Heitzinger, and O. Kimberger, “Optimal timing for renal replacement therapy in critically ill patients using reinforcement learning algorithms,” Journal of Critical Care, vol. 86, p. 154964, 2025.
• M. M. Azarbeik, H. Razavi, K. Merat, and H. Salarieh, “Augmenting inertial motion capture with slam using ekf and srukf data fusion algorithms,” Measurement, vol. 222, p. 113690, 2023.
• A. Meghdari, S. M. J. Zolanvari, H. Izanlo, M. S. Tohidi Nafe, M. M. Azarbeik, and et al., “An overview of the design experience and group analysis of a spinning ride from the perspective of engineering education,” Iranian Journal of Engineering Education, vol. 24, no. 94, pp. 39–60, 2022.

Conference Proceedings

• V. Arzt, M. M. Azarbeik, I. Lasy, T. Kerl, and G. Recski, “TU Wien at SemEval-2024 task 6: Unifying model-agnostic and model-aware techniques for hallucination detection,” in Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), pp. 1183–1196, Association for Computational Linguistics, June 2024.