Dumitru Verșebeniuc

M.Sc. candidate in Artificial Intelligence

Research interests in natural language processing, transformer generalization, retrieval-augmented generation, agentic systems, mathematical optimization and physically grounded video and image generation.

Maastricht, Netherlandsd.versebeniuc@dikaver.com

GitHub LinkedIn Instagram Google Scholar ResearchGate ORCID

Publications

2026OngoingM.Sc. Thesis, Maastricht University

Co-Optimization of Inverter Placement and Cable Routing for Distributed Inverter Topologies in Utility-Scale Solar Plants

byD. Verșebeniuc

supervised byM. Mihalak

Abstract

Formulated the layout design as an Integer Linear Program (ILP) co-optimizing panel-to-string partitioning, inverter placement, and DC/AC cable routing with geometric and electrical constraints; exper...

2025Preprint

From Pattern Recognition to Reasoning: A Survey for Transformer Generalization and Recall in Algorithmic Tasks

byD. Verșebeniuc

supervised byA. Härmä

Abstract

This paper surveys recent advances in enhancing the generalization capabilities of Transformer models in algorithmic tasks, marking a shift from pattern recognition to algorithmic reasoning. We introd...

Transformer GeneralizationAlgorithmic ReasoningLength GeneralizationCompositional GeneralizationPositional EncodingData Formatting Strategies

PDF

2024Accepted at BNAIC/BeNeLearn 2024 · To appear in Springer CCIS

Generative AI-Based Virtual Assistant Using Retrieval-Augmented Generation: An Evaluation Study for Bachelor Projects

byD. Verșebeniuc, M. Elands, S. Falahatkar, C. Magrone, M. Falah

supervised byM. Boussé, A. Härmä

Abstract

Large Language Models have been increasingly employed in the creation of Virtual Assistants due to their ability to generate human-like text and handle complex inquiries. While these models hold great...

Natural Language ProcessingRetrieval-Augmented GenerationInformation RetrievalEducational TechnologyAI Evaluation MetricsInteractive AI

PDF Code Demo Presentation

2023B.Sc. Thesis, Maastricht University

Model-Based Clustering Multivariate EMA Time-Series Data

byD. Verșebeniuc

supervised byJ. Spanakis, M. Ntekouli

Abstract

In recent times, smartphones and watches have facilitated researchers in collecting real-time through ecological momentary assessment (EMA). This data enables the exploration of human behavior, emotio...

Model-Based ClusteringMultivariate Time SeriesEcological Momentary AssessmentMental HealthFeature Extraction

PDF Code

Education

M.Sc. Artificial Intelligence

Maastricht University

Sep 2023 – Present

B.Sc. Data Science & Artificial Intelligence

Maastricht University

Sep 2020 – Jun 2023