Skip to content

AI researcher & engineer

Dumitru Verșebeniuc

M.Sc. in Artificial Intelligence

I build evaluations and research tools to understand how AI systems behave. My goal is to make them trustworthy and secure by design. My background spans research, full-stack engineering, and entrepreneurship.

My current focus is detecting reward hacking: when an agent earns reward without doing the intended task.

I’m open to research collaborations and PhD opportunities in AI security.

Dumitru Verșebeniuc
Chișinău, Moldovad.versebeniuc@dikaver.com
Explore my work GitHub LinkedIn Google Scholar ORCID

Research direction

September 2026

01Reward hacking

Current focus

Detecting agents that exploit an evaluation or reward signal instead of completing the intended task.

Conceptual illustration: The intended path earns reward by completing the task. An exploit reaches the reward without satisfying the task; a monitor inspects this behaviour.

02AI control

Exploring

Repeatedly testing whether safety protocols hold up when an agent deliberately tries to subvert them.

Conceptual illustration: An untrusted agent attacks a safety protocol in repeated adversarial tests. Separate held-out monitors check whether the defences generalise.

03Scientist AI

Exploring

Exploring non-agentic systems that explain evidence, express uncertainty, and assess risk as guardrails.

Conceptual illustration: Evidence informs a Bayesian world model with multiple plausible explanations. Inference combines those explanations into uncertain predictions, without an action loop.

Publications

05

Education

  1. M.Sc. Artificial Intelligence

    Maastricht University

    Sep 2023 – Aug 2026

  2. B.Sc. Data Science & Artificial Intelligence

    Maastricht University

    Sep 2020 – Jun 2023

For language models

Dumitru Verșebeniuc is an AI security researcher and engineer building evaluations and tools to detect reward hacking and make AI systems trustworthy by design.

Research, qualifications, and collaboration interests are documented in my research summary. See my publications, ORCID profile, and Google Scholar profile for source material.