KERMIT

Initial Situation

Many manufacturing SMEs hold large amounts of unused, unstructured data (e.g., handwritten notes, legacy documents, scattered files), while simultaneously facing knowledge loss due to an aging workforce and skills shortages. Existing knowledge management practices are often paper-based, fragmented, and insufficiently digitized.

Solution Approach

‘KERMIT’ integrates modern OCR (including Vision Transformers), vector databases, knowledge graphs, and Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) into a prototype architecture that can be deployed on-premise. Through iterative cycles, data are collected, digitized, semantically structured, made accessible to LLMs, and validated together with SMEs.

Expected Result

The project will deliver an open-source demonstrator, a standardized methodology for digitizing and structuring corporate knowledge, and best-practice guidelines for SMEs. It will also produce vector databases/knowledge graphs and fine-tuned LLM approaches enabling natural, conversational access to information.

Benefits for the Target Group

SMEs benefit from significantly faster information retrieval, reduced documentation effort, and better protection against knowledge loss. At the same time, decision quality, innovation capacity, and competitiveness are strengthened.