Internal ChatGPT with RAG pipeline for the Feddersen Group

The group wanted to build a data protection-compliant and cost-efficient GPT solution: billing on a consumption basis instead of license costs per user, secure provision of internal information in the chat context, rights and client management as well as chat history and feedback loop.

We have developed an individual React frontend that covers and extends the core functions of ChatGPT - for example with direct references to the documents that led to the answer. In the backend, a REST API processes the requests and communicates with Azure. The RAG pipeline dynamically provides the appropriate information. Every component – ​​including the LLM – remains interchangeable.

• Compliance through data protection-compliant solutions and rights management
• Access to internal data of all companies via the RAG pipeline
• Effizienzsteigerung durch GPT-Nutzung in den Fachabteilungen
• Cost reduction through consumption billing instead of licenses per user