Philipp Gordetzki
Philipp Gordetzki is an AI Engineer at IBM Switzerland and Associated Researcher at the University of St. Gallen. He recently defended his doctoral dissertation at the University of St. Gallen. His research on AI for resilient supply chains and AI-human creativity has been presented at international conferences and has been accepted for publication in Information Systems Research. In his work as an AI Engineer, he has designed and built RAG-based pilot systems for clients across industries.
Beitrag
When AI gives wrong answers, the reflex is to reach for a bigger model. This reflex is expensive and energy-hungry, yet it rarely fixes the problem. Most organizational knowledge sits in documents such as guidelines, contracts, and reports, and what is missing is not intelligence but access. This talk presents an open-source architecture for Retrieval-Augmented Generation (RAG) that supports small open models, grounded in documents prepared by Docling, an open-source project developed in Switzerland, and retrieved through hybrid search with OpenSearch. The principle throughout is sufficiency over size: by providing the right knowledge, we do not need ever-bigger models.