PhD Researcher · TU Wien × Med. Uni. Vienna

Mohammad MahdiAzarbeik

Reinforcement learning for the bedside.

I build and evaluate offline reinforcement-learning agents that learn treatment policies from intensive-care data, and test whether those policies hold up against the standards clinicians actually need.

Read the research Scholar LinkedIn
// About

Where the algorithm meets the ward

Mohammad Mahdi Azarbeik
MA

I work on machine learning for critical care, with a focus on reinforcement learning, off-policy evaluation, and — increasingly — large language models. I hold an M.Sc. in Mechanical Engineering and am completing a PhD in Computer Science (Informatics) at Vienna University of Technology.

My research applies RL methods to real clinical data at the Medical University of Vienna, closing the distance between what an algorithm can optimise and what a clinician can trust. The interesting problems live in that gap: cohort definitions, evaluation you can defend, and policies that survive contact with a second hospital.

Research channels
CH1

Clinical AI & decision support

Reinforcement learning for critical care, clinical decision-support systems, and rigorous off-policy evaluation.

CH2

Machine learning

Deep reinforcement learning, generative AI and LLMs, self-supervised representation learning.

CH3

Biomedical data

Physiological time series and multi-modal clinical records (EHR, structured, imaging) from ICU databases.

CH4

Robotics

State estimation, sensor and data fusion, localization.

// Experience

Positions & teaching

Appointments
Jul 2025 — Present
Researcher
Medical University of Vienna (MUW)
Feb 2025 — Apr 2026
Researcher
Ludwig Boltzmann Institute · Digital Health & Patient Safety (LBI DHPS)
Oct 2023 — Jan 2025
University Assistant
Vienna University of Technology (TU Wien)
Teaching
  • →
    Reinforcement Learning
    TU Wien
  • →
    Generative AI
    TU Wien
  • →
    Data-oriented Programming Paradigms
    TU Wien
  • →
    Numerical Analysis
    K. N. Toosi University of Technology
// Publications

Selected work

2026
05
Learning and evaluating improved reinforcement-learning-based policies for sepsis treatment on MIMIC-IV
M. M. Azarbeik et al.
Journal of Critical Care
DOI ↗ PDF
2025
04
Optimal timing for renal replacement therapy in critically ill patients using reinforcement-learning algorithms
L. Kapral, M. M. Azarbeik et al.
Journal of Critical Care
DOI ↗ PDF
2024
03
TU Wien at SemEval-2024 Task 6: Unifying model-agnostic and model-aware techniques for hallucination detection
V. Arzt, M. M. Azarbeik et al.
SemEval-2024 Proceedings · ACL Anthology
DOI ↗ PDF
2023
02
Augmenting inertial motion capture with SLAM using EKF and SRUKF data-fusion algorithms
M. M. Azarbeik et al.
Measurement · Elsevier
DOI ↗ PDF Visual
2022
01
An overview of the design experience and group analysis of a spinning ride from the perspective of engineering education
A. Meghdari et al.
Iranian Journal of Engineering Education
DOI ↗
// Education & toolkit

Academic & technical

Degrees
PhD Candidate · Computer Science
Informatics
Vienna University of Technology (TU Wien)
M.Sc.
Mechanical Engineering
Sharif University of Technology (SUT)
B.Sc.
Mechanical Engineering
K. N. Toosi University of Technology (KNTU)
Core competencies
Programming
PythonMATLABSQLLinux
Clinical data
MIMIC-IVViennaAIdbICU time-series preprocessingEHR feature engineeringPhysiological signals
Deep learning & ML
PyTorchscikit-learn
Robotics
Sensor & data fusionKalman filteringROSArduino
Tools
GitGCPPostgreSQLScientific writing
// Contact

Let's talk research.

Open to collaboration on reinforcement learning, clinical decision support, and off-policy evaluation.

mm.azarbeik@gmail.com Google Scholar