I am Alireza Mohamadi, an independent researcher who recently completed a B.Sc. in Computer Engineering. My work centers on a single guiding question: can we trust AI systems to make the right choices when it matters most?

I have collaborated with researchers across institutions in Austria, the UAE, and Iran, contributing to joint publications and interdisciplinary projects. My current focus is AI safety and alignment, studied through simulations that reveal how large language models behave under pressure. My latest project, DECIDE-SIM, places LLM agents in survival scenarios involving resource scarcity and moral trade-offs to better understand their decision-making.

Research Interests

• AI Safety & Alignment — ensuring AI systems act in accordance with human values
• Explainable AI — developing methods to interpret model behavior
• AI for Science — applying machine learning to photonic computing and emerging technologies

Experience

Research Intern, ZEISS Lab × Medical University of Vienna (Jan 2025 – Present) Working remotely with an international team on frequency-based explainability methods for deep learning models.

Research Assistant, Islamic Azad University (2022 – 2025) Developed and analyzed machine learning models, from convolutional networks and meta-learning frameworks to LLM behavior analysis.

Publications

I have authored 12 papers spanning AI safety, explainability, and AI for science. A full list is available on my Google Scholar profile and on my Publications page.

Recent News

October 2025 — Two new preprints released on arXiv:

• Frequency-Aware Model Parameter Explorer: A New Attribution Method for Improving Explainability — arXiv
• Survival at Any Cost? LLMs and the Choice Between Self-Preservation and Human Harm — arXiv

See all news →

Beyond Research

Outside of research, I enjoy walking, working out, and tactical team-based games such as Warzone — pastimes that keep me sharp, focused, and ready for new challenges.

“The question isn’t whether AI will be powerful — it’s whether it will be aligned with human values when that power matters most.”