Category: Research

Posted on

“Generative AI Intelligent Process Automation Platform” project on Leonardo HPC

Uhura Solutions is developing AI platform for document-driven process automation in financial services. The “Generative AI Intelligent Process Automation Platform – GAIPAP” project is a transformative initiative aimed at revolutionizing the financial industry through the integration of advanced AI-driven automation solutions that combine capabilities of Large Language Model (LLM), low-code development, and process automation workflows. This unique fusion of technologies holds the potential to significantly enhance operational efficiency, cost reduction, and customer experience within the financial sector.

An exceptional aspect of the platform is the incorporation of fine-tuned Large language models (LLMs), which sets it apart from generic AI solutions. These LLMs, refined for the financial industry’s unique language, context, and patterns, promise to deliver more precise and relevant insights. The platform’s value proposition is further strengthened by its ability to streamline workflows, offer scalability, real-time data insights, and elevate customer experiences. The integration of a low-code development framework and process automation workflows enables swift prototyping and deployment of AI-driven models, thereby accelerating time-to-market and capitalizing on market opportunities effectively.

The project involves the use of open-source Python libraries and pre-trained Large language models which are fine-tuned on a custom-made private dataset created by the team, allowing for specialized tasks within the financial sector. Leonardo supercomputer enabled us to make a leap with GPU training experiments from 1B to 7B and higher Large language models. We are focusing on various LLM research and optimization methods in this phase of the project, which include hyperparameter tuning, model quantization and pruning, and parameter efficient fine-tuning (PEFT). Special attention is put on dataset preprocessing including quality filtering, deduplication and privacy redacting.

This project, developed and managed by Uhura Solutions, was awarded with 4,500 node hours of GPU resources (8x64GB) on the Leonardo Buster HPC for a duration of 12 months.

Posted on

Master thesis: HPC/AI for breast cancer detection

Ms. Tamara Pavlovic defended her MSc thesis on the use of HPC/AI for creating prediction models for breast cancer detection on 23.10.2024. With the support from NCC Montenegro, Ms Pavlovic did her research in the context of the HPC4S3ME project and the focus was on AI and computer vision applications in medicine. From the motivational point of view, we congratulate Tamara for finalizing and defending her thesis during the Breast Cancer Awareness Month (‘Pink October’) as people around the world adopt the pink colour and display a pink ribbon to raise awareness about breast health.

ABSTRACT – Artificial Intelligence (AI) is revolutionizing numerous sectors, including medicine, by offering innovative methods for diagnosing, treating, and researching diseases. This master’s thesis focuses on the application of AI in the diagnosis of breast cancer, using computer vision algorithms to analyze mammographic images. Through a combination of convolutional neural networks (CNNs) and deep learning, models have been developed that identify malignant changes, potentially contributing to earlier and more precise disease detection. The thesis examines in detail how AI can improve the efficiency of screening processes, reduce the time required for diagnosis, and enable a more personalized approach to treatment. In addition to technological progress, ethical issues such as patient safety and the transparency of AI systems are also considered. The results of this study confirm that the application of AI in breast cancer diagnostics can significantly enhance medical procedures. The models tested, ResNet152 and DenseNet121, demonstrated quite good performance in classifying breast cancer. Their AUC scores, which exceed the threshold of 0.9, indicate their potential for use in clinical practice. These findings not only contribute to the improvement of diagnostic processes but also open up opportunities for further research and development of AI technologies in medicine.

This research was done in th context of HPC4S3ME and with the support from EUROCC NCC Montenegro
Ms Pavlovic finalized her thesis during the Breast Cancer Awareness Month (‘Pink October’)
Posted on

Master thesis: HPC/AI in precision agriculture

Mr Mato Martinovic defended his MSc thesis on 23 octiber 2024. His research focused on detecting plant deseases for applications in vineyards. He was experimenting with HPC/AI and computer vision. He is one of the latest graduates from the AI master program created under EUROCC project and his mentoring was done with the support of EUROCC NCC Montenegro.

ABSTRACT – This research analyzes the use of computer vision in the field of viticulture. The thesis describes problems in viticulture, computer vision and its use in this field. The paper analyzed the performance of ResNet50, VGG16 and MobileNet models in the classification of diseases and grapevine species. The models achieved accuracy of 98.67%, 97.28%, and 98.72% on the original test data set, while on the extended one, they achieved 87.47%, 72.07%, and 86.64%, respectively, when classifying diseases. In species classification, the models achieved accuracies of 70%, 78% and 88% on the original test data set, and 66%, 51% and 72% on the extended one, respectively. The VGG-16 model had the largest difference in accuracy over extended data, while ResNet had the smallest decrease in accuracy in both cases, which implies that ResNet generalizes the data better. The paper presents the process of creating a platform that allows users to post an image and receive a prediction value through a mobile application.

HPC/AI and computer vision for applications in smart viticulture
Posted on

Master thesis: AI/ML and applications in medicine

Mr. Luka Jeremic defended his MSc thesis on 23 October 2024. The title of the thesis was AI and applications in medicine. His research was mentored by HPC4S3ME team members and it was done in the context of AI master program at the Faculty for information systems and technologie at UDG. This program and Master students are supporter by EUROCC NCC Montenegro.

ABSTRACT – This research explores the application of artificial intelligence in medicine, with a focus on the classification of brain, liver, and blood cell diseases. The main objective is to evaluate the effectiveness of algorithms in recognizing and classifying diseases of these organs. Through the development of a prototype information system, the study analyzes how artificial intelligence can improve diagnostics and contribute to the advancement of personalized medicine. The methodology includes a literature review, the development of computer vision models, and the assessment of model accuracy using real medical data. The results show that models based on deep neural networks can enhance the accuracy and speed of diagnostics, allowing for more precise disease classification. The paper also highlights the barriers and challenges in implementing these technologies,
including the need for ethical considerations and training of medical staff. The conclusions suggest that this approach has the potential to significantly improve medicine, but further research and refinement are necessary.

Mr Jeremic defended his master thesis on AI/ML and applications in medicine
Posted on

Master thesis: Deep learning in energy sector

Ms. Zoja Scekic, a young researcher on HPC4S3ME project, defended her MSc thesis “Deep learning and applications in energy sector” today. This is one of the main project outputs in capacity building aimed at HPC/AI skills for applications in priority domains of Montenegrin S3.

ABSTRACT – This master’s thesis explores the application of advanced deep learning models for predicting day-ahead electricity prices, focusing on the accuracy and efficiency of these models compared to traditional forecasting methods. With the increasing integration of renewable energy sources and the growing complexity of electricity markets, accurate price forecasting has become crucial for market participants, grid operators, and policymakers. The research is structured around four case studies, each employing different deep learning techniques, such as Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and hybrid models like CNN-LSTM. Despite the promising results, the research recognizes limitations related to data quality, model complexity, and computational resource requirements. The study emphasizes the need for further research into optimizing model efficiency, integrating more diverse data sources, and expanding the applicability of these models to different energy markets.

Ms Zoja Scekic defended her MSc thesis on Deep leaning applications in energy sector
This MSc thesis was done in the context of HPC4S3ME with support from EUROCC NCC Montenegro
Posted on

In-house HPC lab infrastructure update

As planned, our project AI-AGE is advancing high-performance computing (HPC) infrastructure to support AI-driven research on biomarkers of aging in medical applications. This initiative will empower our team with cutting-edge resources, allowing us to enhance our capacity for data analysis and predictive modeling. To meet the demands of sophisticated AI computations, with the support of AI-AGE, we are upgrading our existing HPC setup with a powerful computing node. This new addition includes a rack computing node equipped with a 48 CPU cores with 128GB RAM, NVIDIA L40 48GB GPUs, and 2x480GB internal SSDs. In addition, the project supported NAS storage of 24TB (multiple disks with RAID) dedicated for dataset management. This infrastructure enhancement is designed to integrate smoothly with our existing equipment, augmenting both our computational and storage capabilities while providing significant value for our investment.

New computing infrastructure supported by the AI-AGE project as planned

AI-AGE project, supported by the Ministry of education, science and innovation, is implemendet through collaboration between Faculty for information systems and technologies at Uiversity of Donja Gorica, and Faculty of medicine at University of Montenegro. The in-house HPC infrastructure is a result of cross-project collaboration with HPC4S3ME project (IPA programme) and both of these project are done with the support from EUROCC NCC Montenegro. The main goal for the in-house lab is for researchers to gain a hands on experience with physical equipment a their disposal, while for larger computing tasks, we will apply for computing time on some of the EU supercomputers.

Click on image to open AI-AGE project website
Posted on

Cross-NCCs training event: “Machine Learning for Multiple Domains”

NCC Türkiye, NCC Serbia, NCC Montenegro, and NCC North Macedonia are pleased to announce a joint online training event titled “Machine Learning for Multiple Domains: From Concepts to Implementation,” scheduled for October 14-15, 2025, starting at 10:00 AM.

Participants can expect engaging program at both beginner and intermediate levels and will feature hands-on sessions along with key presentations spread across two days: Brief introductions on HPC/AI activities by 4 NCCs; Supercomputing access demo (TRUBA HPC infrastructure), presentations from NCC Macedonia (Design, Develop, Deploy, and Iterate on Production-Grade ML Applications) and NCC Turkey (Protein Language Models and Using Them for Downstream Prediction Tasks) on Day 1, and presentations from NCC Serbia (Modeling of Large-Scale Social Data) and NCC Montenegro (Analyzing Social Media Trends) on Day 2.

We look forward to your participation in this valuable opportunity to enhance your skills in HPC and machine learning!

More information on the training program, timetable and registration details you can find here: https://indico.truba.gov.tr/e/ML4MultipleDomains

Privacy Policy Proudly powered by WordPress