The role of UL FE in the Slovenian AI Factory (SLAIF)

The Slovenian AI Factory (SLAIF) has been launched in Slovenia as a national infrastructure intended for companies, research organizations and the public sector for the development and use of advanced artificial intelligence (AI) solutions. The project represents an important step towards strengthening the country’s technological independence and competitiveness in this rapidly growing field.

Within the project, the Faculty of Electrical Engineering of the University of Ljubljana (UL FE) plays an important role, researchers from the Laboratory for Machine Intelligence, the Laboratory for Robotics and the Laboratory of Control Systems and Cybernetics are involved in delivering key services. These are designed as comprehensive workflows that enable users to move efficiently from an initial idea to a working solution. Training in the field of artificial intelligence will also be provided for different user groups, ranging from industry to the public sector.

Across Europe, AI factories are emerging as structured ecosystems that help organizations turn AI capabilities into practical and repeatable results. The focus is on supporting end users through a combination of scalable computing resources, data processing, reusable workflows and tools, as well as training and direct expert support. The goal is not only access to infrastructure, but also reliable and predictable implementation of AI solutions. SLAIF brings this concept to the national level by coordinating partners, resources and activities into a unified offering. It connects concrete use cases with implementable workflows and supports their adoption through services and training.

The role of UL FE

UL FE contributes to SLAIF through the development and implementation of services designed as workflows, as well as through training. The focus is on practical solutions and training that enable users to move from initial interest to responsible use of AI in practice.

Worksflows

UL FE offers three services:

• AI-supported workflow for grading scanned handwritten exams in science and engineering (STEM)

This solution supports the grading of scanned handwritten exams, including diagrams. It is designed to reduce repetitive work in grading while maintaining full control for the evaluator. The system generates structured grading records and flags cases that require quick human review.

• Initial preparation of datasets from unstructured visual data using foundation models (automatic annotation)

A configurable workflow enables the transformation of large volumes of existing images or videos into annotated datasets ready for model training. The focus is on extracting labels from captured data (without generating synthetic content), standardized output formats and quality control logs that accelerate the deployment of computer vision solutions.

• Obstacle detection for autonomous vessels and vehicles based on anomaly detection

This workflow enables the detection of obstacles in the surroundings of vessels and vehicles within limited geographic areas. It is intended for rapid prototyping in cases where large annotated datasets are not available.

The first two services are expected to be available from 2027, and the third from 2028.

Training

UL FE also provides two training programmes tailored to user needs and the principles of responsible AI adoption:

• Introducing AI into teaching

This practical workshop is intended for educators in primary, secondary and higher education, as well as training providers. It addresses common uses of AI in teaching and principles of responsible use. It includes an overview of the user journey of a prototype SLAIF workflow for automated assessment, as well as guidance on using AI-supported learning tools.

• Practical image formation and data acquisition for computer vision

This training covers the basics of image formation and acquisition, common pitfalls in ensuring data quality, and basic guidelines for preparing image data for use in computer vision (without model training).

The first training is planned for autumn 2026, and the second for 2027.

These activities reflect the current development plan and will be adapted in the future to user needs and partner contributions.

More about SLAIF and the role of UL FE was presented by Assoc. Prof. Dr Janez Perš on the First Programme of Radio Slovenia.