Mr. Elvis Taruh successfully defended his master’s thesis titled “Development of Edge/AI Applications with HPC” at the Faculty of Information Systems and Technologies, University of Donja Gorica.
Mr Elvis Taruh
ABSTRACT – The efficiency of training artificial intelligence (AI) models has become a crucial factor in modern research, especially when dealing with complex systems that require substanial computational power. This study explores how the application of high-performance computing (HPC) and Edge devices can optimize the AI model training process, reducing processing time and improving efficiency. Through an experimental approach, AI model training was analyzed across three different platforms. Local computer, Google Colab and the HPC cluster at the University of Donja Gorica. As a practical example, livestock detection was used. By comparing the training time, memory consumption, and model accuracy, the research demonstrates that HPC clusters significantly accelerate the training process compared to traditional methods, while Edge devices enable faster real-time data analysis.
There was around 30 people attending. This was a small celebration for EuroCC2 and EuroCC4SEE projects
Mr. Arnad Lekić successfully defended his master’s thesis titled “Development of an AI Tutor Using Large Language Models and HPC” at the Faculty of Information Systems and Technologies, University of Donja Gorica.
Mr Arnad Lekic
ABSTRACT – This thesis explores the development of a personalized AI tutor using large language models (LLMs), with a specific focus on the LLaMA architecture and the application of High-Performance Computing (HPC) resources. The research involves the acquisition, setup, and evaluation of an open-source LLaMA model, with the goal of building a system capable of automated test grading. Special emphasis is placed on the training efficiency and feasibility of running the model locally using the available computing nodes, compared to cloud-based solutions like Google Colab. Beyond the technical implementation, the study also addresses the ethical challenges of using generative AI in education. Through experimental analysis, the research demonstrates that open models can be effectively adapted for educational purposes, with the potential to expand to grading diverse exam formats and generating educational content. The work provides directions for future development of systems leveraging advanced multimodal models for more complex tasks.
The defence was attended by over 30 people. We had three candidates that day, all in the context of EuroCC2 and EuroCC4SEE
Ms. Enisa Trubljanin successfully defended her master’s thesis titled “Deep Learning with Application in Education” at the Faculty of Information Systems and Technologies, University of Donja Gorica. The development and testing of these solutions were supported by high-performance computing (HPC) resources provided through the EuroCC initiative in Montenegro.
Ms. Enisa Trubljanin
ABSTRACT – This master’s thesis explores the potential application of deep learning in education through the development and evaluation of two concrete solutions: an intelligent chatbot for solving matrix problems and a model for detecting cheating during online exams by analyzing eye movement. The first part of the thesis provides a theoretical foundation of deep learning, with a focus on neural networks, their architectures, transfer learning, and evaluation metrics. The practical part presents the development of a chatbot based on advanced language and mathematical models, implemented using high-performance computing cluster resources, enabling students to engage in interactive mathematics learning. Additionally, a model for detecting cheating through gaze analysis was developed, trained on the Columbia Gaze Dataset, and integrated into an online exam proctoring system. Evaluation results demonstrate a high level of accuracy and user satisfaction for both solutions. Beyond the technical aspects, the thesis also addresses ethical issues and privacy concerns related to the use of artificial intelligence in educational settings. Based on the findings, the study highlights the broad range of potential applications of deep learning in modern educational systems.
We are pleased to announce that two Montenegrin SMEs, Recrewty and DigitalSmart, have been awarded access to the Leonardo Booster at CINECA, one of Europe’s most powerful high-performance computing (HPC) systems. The access is granted under the EuroHPC JU Development Call, providing 12 months of HPC resources to support advanced AI development and innovation.
The awarded project, HPC4HR – High-Performance Computing for Human Resources, aims to revolutionize recruitment processes in the Western Balkans through the integration of generative AI (GenAI) and HPC technologies. The project focuses on analyzing multimodal data—text, audio, and video—from candidate applications to enable efficient, unbiased, and culturally sensitive hiring processes. The HPC resources will be used in support to their joint FFPlus project GenAI-HPC4WB (link).
The GenAI-HPC4WB project combines GenAI, ML, and HPC to optimize hiring processes and enhance business culture in the Balkan region
To achieve this, the project will leverage cutting-edge AI models such as LLaMA 3.x, Mistral, Wav2Vec 2.x, Whisper, FaceNet, and DeepFace, all fine-tuned using HPC resources at CINECA. These models will be developed to evaluate resumes, analyze psychometric traits from audio recordings, and interpret emotional and behavioral cues from HR data. By utilizing HPC, the project ensures scalability and processing speed, enabling large datasets to be handled efficiently, significantly shortening recruitment cycles and improving candidate matching. In addition to enhancing operational efficiency, HPC4HR emphasizes ethical AI development, aiming to reduce biases in recruitment and promote diversity and inclusion. The expected outcomes include a set of adaptable, high-performance AI tools that set new standards for digital transformation in human resources, applicable across industries and geographies.
Leonardo Booster partition will be used in this project
The success of Recrewty and DigitalSmart reflects the growing impact of NCC Montenegro in strengthening the national HPC and AI ecosystem. By supporting companies in accessing leading European HPC infrastructure, NCC Montenegro continues to drive digital innovation, competitiveness, and technological excellence in the region.
The Faculty for Information Systems and Technologies (FIST) at the University of Donja Gorica (UDG) has been awarded a prestigious grant through the EuroHPC Joint Undertaking Open Call, marking a significant milestone for Montenegrin academic engagement with cutting-edge high-performance computing (HPC) resources.
As part of this grant, FIST has secured access to the Leonardo Booster partition at CINECA, one of the most powerful supercomputers in Europe. This will enable FIST researchers to perform large-scale experiments that are otherwise infeasible with standard computing infrastructure.
FIST at UDG gained access to Leonardo BOOSTER via EuroHPC JU open calls
The awarded project focuses on cross-lingual transfer learning in large language models (LLMs), aiming to systematically evaluate how model architecture and scale influence multilingual performance. By fine-tuning major LLM families (LLaMA, Mistral, DeepSeek) across model sizes from 1B to 70B parameters, the research will generate insights into optimal model selection under real-world resource constraints—critical for European institutions working with diverse languages and limited compute budgets. This is a development project with HPC resources available for 12 months.
The research focuses on cross-lingual transfer learning in large language models (LLMs)
This achievement underscores the growing capacity of UDG and FIST to contribute to frontier AI research, while reinforcing the mission of the National Competence Center in HPC (NCC Montenegro) to support HPC adoption across academia and industry in the region.
We congratulate the FIST team on this major success and look forward to sharing results from their HPC-powered investigations.
As part of a broader effort to analyze operational patterns in Montenegro’s healthcare sector, a predictive model was developed to estimate indicators related to the quality of healthcare services provided to citizens at the secondary and tertiary levels. The model was initially trained on a sample of 10,000 records and later expanded to over 40,000 records using the HPC cluster at UDG, which enabled the training process to complete in under one minute. All records were fully anonymized in accordance with ethical standards, and the dataset incorporated a wide range of temporal, demographic, and institutional variables relevant to healthcare access and service delivery patterns.
Using AI and HPC to analyze operational patterns in Montenegro’s healthcare sector
The model was implemented using the Random Forest Regressor algorithm, and additional experiments were conducted with alternative modeling techniques and target transformations. The baseline configuration achieved strong predictive performance (MAE ≈ 95 hours, R² = 0.87). While log-transformation and alternative algorithms such as XGBoost were evaluated, they did not yield improvements over the initial approach. Feature importance analysis revealed that factors such as month, weekday, and clinical unit had the highest impact on model outputs. Access to dedicated computing resources has been granted for further development, enabling continued model training and experimentation with advanced algorithms and optimization methods.
Click on image to learn more about Exploring – an innovative software development SME from Montenegro
The paper “Transforming Matrix Problem Solving with Intelligent Tutoring Systems” by E. Trubljanin, E. Taruh, S. Cakic, T. Popovic and L. Filipovic was presented at INFOTEH conference an is published by IEEE Xplore. Researchers from the UDG with support from HPC NCC Montenegro have developed an innovative intelligent tutoring system that leverages artificial intelligence (AI) and high-performance computing (HPC) to revolutionize how students learn matrix operations. This chatbot-based solution combines Optical Character Recognition (EasyOCR) with an advanced natural language processing model (Qwen2-Math-7B-Instruct) to interpret both text and image inputs, enabling it to perform matrix operations such as transposition, addition, and multiplication while providing clear, step-by-step explanations. Supported by the university’s HPC infrastructure, the system ensures high-speed processing and real-time feedback, achieving up to 99% accuracy in matrix recognition from high-quality images. Designed with education in mind, this AI-powered tutor enhances interactivity, understanding, and learning outcomes for students tackling complex linear algebra concepts, and sets the stage for future enhancements like handwritten input recognition and support for more advanced operations.
ABSTRACT – This paper presents the integration of optical character recognition (OCR) and advanced natural language processing (NLP) models for automated handling of matrices derived from images and textual inputs, all combined within an implemented chatbot. The motivation for choosing this topic arises from the practical experiences of the authors gained while working with groups of students who encounter the concept of matrices as part of their academic responsibilities. Through the analysis of their results and classroom interactions, it was observed that many students struggle with this area. This paper presents an innovative approach to enhancing matrix problem-solving by leveraging intelligent tutoring systems supported by High-Performance Computing, aiming to improve learning efficiency and student outcomes. By combining the EasyOCR framework and the Qwen2-Math-7B-Instruct model, operations such as transposition, addition, and multiplication of matrices are enabled. The system supports the input of one or two matrices, allowing the selection of operations through textual or image-based queries. The OCR component extracts numerical data from images, while the NLP model interprets user requests and executes operations accurately. The interface allows the addition of a second matrix image only when necessary, enhancing the system’s intuitiveness and efficiency. The results of the recognition accuracy of the OCR model of image input matrices of different dimensions show a high level of accuracy of 95%, while for 2×2 matrices they reach an accuracy of 99%. This work contributes to the development of AI-powered tools for mathematical operations and holds potential applications in education.
