Generative AI for SMEs in Saxony

Making better use of digitalization potentials through the GenAIsys application center

Small and medium-sized enterprises in particular often find it difficult to exploit the enormous potential of digitalization and automation because the costs are too high and they lack expertise. This is where the GenAIsys application center at Fraunhofer IVI comes in: it provides SMEs with the necessary technical infrastructure and takes on the development and training of generative AI solutions tailored specifically to the respective companies. Through the intelligent and secure use of company data, it is possible, for example, to accelerate processing procedures, configure assistance systems specifically for each company, or identify innovative material alternatives.

GenAISys attaches great importance to transferability: flexible, standardized, and reusable software modules are created that are designed to be quickly customized to the needs of a wide range of industries and applications.

Prototype implementation in use cases across various industries

At the heart of the project is the prototypical implementation of specific use cases in collaboration with SMEs from Saxony. The associated corporate partners come from various industries, such as logistics, transportation, and construction. They will apply and test the results within the framework of the project. It is important to note here that they will retain sovereignty over their data in all cases. This will enable them to benefit from digitalization securely and with relatively little effort, and to increase their competitiveness in the long term.

The use cases utilize both company-owned data and publicly available data. With the help of embedding or retrieval augmented generation (RAG), documents or entire databases become the basis for learning and are used either to fine-tune already trained large language models (LLM) or to train new transformers for the respective use case.

Once the solutions for the use cases have been successfully implemented, the modules are generalized. The goal is to create as comprehensive a library as possible, laying the foundation for the rapid implementation of further future AI use cases – from data preparation to the provision of a wide variety of language models.

Advantages of digitalization for business and administration

Through its development of a robust infrastructure and the establishment of best practices in the field of generative AI, the GenAISys application center is a pioneering project for digital innovation in Saxony's SMEs. In addition, the services developed within the project are also available to public authorities and institutions to leverage their previously untapped potential.

Generative AI, which is at the heart of the GenAIsys project, is one of the key technologies in Saxony's innovation strategy. The project promotes regional expertise in a globally significant field of technology and thus supports the expansion of technological competence in the Free State.

GenAIsys is funded by the Sächsische Aufbaubank - Förderbank with funds from the European Regional Development Fund (ERDF).

The GenAISys use cases

The use case partners are integrated into the project as associated partners. All development steps, beginning with data preparation and ending with the implementation of the modules, are carried out by Fraunhofer IVI umgesetzt. This configuration serves to test the extent to which costs can be kept low for SMEs while still allowing generative AI to be integrated into business processes.

Production

Digital commission documents

  • Automated processing of new orders with the help of a transformer
  • Standards-compliant processing and integration into inventory management system
  • Reduction of manual tasks and error rates with the aim of stabilizing production processes

Logistics

Training assistant

  • Optimal support for knowledge acquisition in continuing vocational education and training
  • Aadaptation of specific knowledge requirements to individual users
  • Training of a language model based on learning content as well as enterprise and user data
  • Reduction of manual tasks and optimization of learning success

Materials science

Virtual materials optimization

  • Development of a trained model able to indicate the potentials of alternative and innovative compound materials
  • Material transformer for data fusion: material composition and physical properties from inspection and testing data
  • Identification of alternative materials with the aim of being prepared in case of raw material shortages, geostrategic challenges, and supply shortages 

Construction industry

Building standards integrator

  • Efficient planning of buildings and infrastructure including automated check of compliance with current standards
  • Acceleration of approval and implementation planning
  • Training of a transformer using relevant standards and internal project data interpreting the standards

Transportation

Environment analyst

  • Establishment of a connection between all road users and the infrastructure 
  • Provision of a realistic image of actual surroundings for the benefit of road safety
  • Processing of traffic surveillance data as well as existing, institute-owned TASC and GIDAS data
  • Utilization for traffic planning and testing of functions for automated driving