ITEA is the Eureka Cluster on software innovation
ITEA is the Eureka Cluster on software innovation

ELFMo

Engineering Large Foundational Models for Enterprise Integration

Project description

Even though it's already happening on a massive scale, integrating Generative AI (GenAI) and Large Foundation Models (LFMs), e.g., OpenAI GPT-x, Google BARD, and FALCON, into existing business and enterprise applications is fraught with high risks and challenges. Considering the huge cost and resources that are required, realizing use cases and applications for which actual added value is created, is already difficult. In addition, there is a larger set of compliance and integration issues, such as a lack of data security, insufficient legal penetration of property rights and IP, unintended bias and lack of fairness, susceptibility to errors, dealing with uncertainties, and lack of transparency. Finally, it is not clear how effective fine tuning, integration and interoperability, continuous training, and maintenance, as well as the continuous supervision and audit could be organized and carried out in an effective way. ELFMo will develop a framework for the reliable integration of LFMs into business applications. Based on the state of practice in industrial software engineering and ML, the project will provide methods, techniques and tools that allow for a risk-based engineering and integration approach that allow for a trustworthy integration of commercial LFMs and fosters the use of open-source models to maintain independence, trust, and transparency. The framework will facilitate enterprise adoption of LFMs in a manner that aligns with European AI regulations, e.g., GDPR and AI Act. ELFMo empowers enterprises to reliably integrate LFMs into software infrastructures and applications by maintaining control over the multitude of risks, challenges and opportunities that are associated with large AI.

Project publications