Publications Archives - HPC Montenegro

03.03.2026.18.03.2026.

Conference paper at IEEE IT2026 on intepretable ML for diabetes screening

AI-AGE team presented a paper titled “Interpretable ML for Diabetes and Prediabetes Screening Using Self-Reported Health Indicators” by S. Lazic, S. Cakic, I. Rubezic Lukic, N. Popovic, and T. Popovic at the 30. Annual Conferenc on Information Technology IT 2026. This was part of mentoring activities and efforts related to development of young researchers.

ABSTRACT – Early identification of type 2 diabetes (T2D) and prediabetes enables timely interventions, yet screening often relies on self-reported data rather than laboratory testing. This work compares lightweight Machine Learning (ML) models: Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Multilayer Perceptron (MLP) trained on 21 self-reported indicators from the 2015 Behavioral Risk Factor Surveillance System (BRFSS) dataset for three-class classification (no diabetes, prediabetes, diabetes). We propose a screening-oriented evaluation where a probability threshold is selected to achieve a target sensitivity (recall) of 0.80. LightGBM achieves balanced accuracy of 0.52 and precision of 0.33 at the target sensitivity, with 38% of cases flagged. Tree SHapley Additive exPlanations (TreeSHAP) highlight general health status, age category, body mass index (BMI), and hypertension as dominant predictors. A FastAPI web application provides individual risk estimates and instance-level explanations. The pipeline demonstrates feasibility of interpretable, calibrated screening from non-laboratory data.

02.03.2026.18.03.2026.

AI and HPC for Honey Authenticity: PollenTrace at IEEE IT2026

At the IEEE IT2026 conference in Žabljak, researchers from the University of Donja Gorica presented PollenTrace, an innovative project combining Artificial Intelligence and High Performance Computing (HPC) to enhance honey authenticity verification. Traditional pollen analysis (melissopalynology), while reliable, is time-consuming and dependent on expert knowledge. PollenTrace addresses this limitation by developing a large-scale microscopy dataset and an AI-driven detection pipeline capable of automatically identifying pollen grains in honey samples.

The project is building a dataset of over 33,000 high-resolution microscopy images derived from more than 1,100 biological samples collected across Montenegro, enabling the development of robust and scalable AI models. As a proof of concept, a deep learning model based on YOLOv11 was trained on annotated microscopy images, achieving 84% precision and 88% recall, demonstrating strong potential for automated pollen detection and future large-scale deployment.

HPC resources played a key role in enabling efficient model training and handling of high-resolution image datasets, highlighting the importance of national HPC infrastructure—such as that provided through NCC Montenegro -in supporting advanced AI applications in agri-food systems. This is also cross-project collaboration.

PollenTrace represents a step forward toward digital, scalable, and reproducible food authenticity verification, with strong potential to support laboratories, regulatory bodies, and industry in ensuring product quality and consumer trust. PollenTrace is supported as a PoC project by the Innovation Fund of Montenegro.

25.02.2026.17.03.2026.

PhD Defence at UDG: Advancing AI and HPC in Precision Agriculture

The University of Donja Gorica, through the Faculty for Information Systems and Technologies, proudly announces the successful PhD defence of Mr. Stevan Čakić, focused on the application of Artificial Intelligence and High-Performance Computing in precision agriculture.

The research addresses key challenges in modern agriculture, particularly in poultry farming, by leveraging deep learning and computer vision models for real-time monitoring, early disease detection, and improved farm management. The models were developed and trained using HPC resources, enabling efficient experimentation and achieving high prediction accuracy exceeding 92% . A significant contribution of this work lies in the integration of HPC-based model development with deployment on edge devices in real farm environments, demonstrating a complete AI-to-industry pipeline. The research also explores the use of generative AI and synthetic data to reduce dependency on large annotated datasets, accelerating innovation cycles.

mr Stevan Cakic presenting his PhD Thesis on AI/HPC in precision agriculture

Importantly, part of this research was conducted in synergy with the FFplus experiment and in direct collaboration with industry partners, highlighting the role of HPC in enabling real-world, industry-driven AI applications. This achievement further demonstrates the impact of the NCC Montenegro and EuroCC2 & EuroCC4SEE initiatives in supporting advanced research, fostering academia-industry collaboration, and promoting the adoption of HPC technologies in strategic sectors such as agriculture.

23.02.2026.25.02.2026.

Researchers from the Faculty of Science and Mathematics published a journal paper on models tested on Leonardo HPC

We are pleased to announce that the research team from the Faculty of Science and Mathematics has published a scientific paper titled “Data augmentation for fuselage panel inspection via 3D point cloud segmentation” in the Journal of Electronic Imaging. The paper presents advanced data augmentation methods to improve fuselage panel inspection using 3D point cloud segmentation, contributing to more accurate and reliable AI-based inspection systems. The research was enabled by access to the Leonardo HPC supercomputing resources, granted through the EuroCC2 project, which allowed the team to process large datasets and develop high-performance models efficiently. More info at: https://doi.org/10.1117/1.JEI.35.3.031202

17.02.2026.17.02.2026.

PAID MNE Showcased in the EuroCC2/EuroCC4SEE Success Stories Booklet — Powering Smarter Trading with Supercomputing

We are proud to highlight PAID MNE as a featured success story in the EuroCC2 & EuroCC4SEE Booklet — demonstrating how HPC is transforming financial analytics and algorithmic trading.

At the heart of PAID MNE innovations lies PAID-T (Price Action Intelligent Detection Trading) — a smart trading platform that leverages advanced algorithms and AI/ML to dynamically adapt to market movements, optimise investment strategies, and manage risk with higher precision. Traditional computing systems quickly reached their limits. To unlock the required performance, the team scaled their solution to the LUMI supercomputer, one of Europe’s most powerful HPC infrastructures. By enabling multinode execution and real-time task distribution, PAID MNE achieved over 1.2 million simulations in under 5 hours — a process that previously would have taken days. This acceleration enables processing billions of historical transactions in hours instead of days, rapid identification of critical market patterns and data-driven optimisation/ increased accuracy of trading strategies.

This achievement, showcased through the EuroCC2/EuroCC4SEE project, demonstrates how supercomputing is becoming a powerful enabler of business innovation. PAID MNE’s journey is a clear example of how HPC and AI together can transform complex, critical data into faster, more profitable decisions.

22.12.2025.29.12.2025.

EuroCC4SEE Forum on HPC/AI-Enabled Business Innovation & PoC Demonstrations

Short video from the event

On 13–14 December, NCC Montenegro hosted the EuroCC4SEE Forum dedicated to business innovation supported by High-Performance Computing (HPC) and Artificial Intelligence (AI). The event brought together representatives from academia, industry, the public sector, and the startup ecosystem, with the aim of strengthening national capacities and supporting Montenegro’s digital transformation.

Welcome note from NCC Montenegro – prof. Bozo Krstajic

We organized networking activities during the breaks

During the two-day program, participants explored practical applications of HPC and AI through Proof-of-Concept demonstrations in the fields of energy, agriculture, health, and mobility, alongside discussions on innovation policies, MLOps approaches, skills development, and business–academia collaboration.

Discussions were focused on AI/HPC driven innovation in business and PoC demonstrators

PoC presentations were based on acamia-industry collaboration

The forum also highlighted the importance of regional and European cooperation, EU funding opportunities, and future activities within the EuroCC4SEE network, contributing to the strengthening of Montenegro’s position within the regional HPC/AI ecosystem. A proceedings booklet with brief elaborations of all covered topics will be made available on the project website.

26.10.2025.24.11.2025.

MSc Thesis on Cross-Lingual Transfer Learning in Large Language Models

Mr. Igor Ćulafić successfully defended his master’s thesis titled “Cross-lingual Transfer Learning in Large Language Models: Scaling Laws and Parameter-Efficient Fine-Tuning for Multilingual Applications.” His research provides a comprehensive study of cross-lingual transfer for the Montenegrin language, combining a custom V-shaped semi-automated book scanner, a YOLOv11 + Tesseract OCR pipeline, and the creation of 46,661 parallel paragraph pairs. Using LoRA fine-tuning on Qwen2.5-7B and Qwen3-30B—executed on the Leonardo EuroHPC supercomputer—the work demonstrates parameter-efficient adaptation (only 1.05% trainable parameters) and offers insights into model behavior in cultural understanding, script mixing, and analytical reasoning. This research was supported by NCC Montenegro team and made use of the HPC cluster and EuroHPC JU computational resources.

V-shaped book scanner prototype used to create datasets

ABSTRACT – This thesis presents a comprehensive study of Cross-lingual transfer learning in Large Language Models with a focus on parameter-efficient fine-tuning for the Montenegrinlanguage. The research integrates the development of a custom semi-automated book scanner with V-shaped design and a computer vision pipeline using YOLO v11 models and Tesseract OCR to digitize 5000 on Montenegrin and 40000 on English language, from public domain books, resulting in 46661 parallel paragraph pairs. Implementation of LoRA fine-tuning on Qwen2.5-7B and Qwen3-30B models was conducted on Leonardo HPC supercomputer, achieving memory efficiency with only 1.05% trainable parameters. Comparative analysis through a structured benchmark of ten progressively complex questions reveals limited but positive effects of fine-tuning, where larger models show better performance in cultural understanding and analytical tasks, while systematic analysis identifies specific problems such as script mixing and cultural inaccuracies that require specialized approaches.