BayVFP-Project KIWIS

The project is a collaboration between the BMW Group and the University of Bamberg, aimed at developing an AI-driven knowledge management system that integrates agentic large language models (LLMs) with modular AI systems. The goal is to improve knowledge access for employees by combining various data sources, while ensuring factual consistency and reducing hallucinations.

By leveraging modular AI models, the agentic LLM activates relevant knowledge modules based on user queries using topic modeling and retrieval-augmented generation (RAG). This modular approach enables efficient processing of different data types and representations, delivering accurate and context-based answers.

The University of Bamberg focuses on reducing hallucinations in LLMs, developing methods to activate AI modules using topic modeling and RAG, and applying cognitive science approaches to design explainable results. The project explores use cases in electric vehicle manufacturing and market analysis, utilizing a chat-based dashboard to facilitate natural interaction with the knowledge base.

The aim is to create a user-friendly and efficient system that enhances transparency and trust, improving knowledge access and interaction for employees in the fields of manufacturing and market analysis.