GPU Servers for AI Inference, Network Storage System and Web Server

Purchase of brand-new IT hardware in three lots Lot 1 GPU servers for AI inference Lot 2 Network storage system Lot 3 Web server The subject of Lot 1 is a dedicated GPU server as an inference backend (of the) RAG system described in Section 2. The server is to be procured partially equipped and must be upgradeable to full configuration with eight graphics processors without replacing core components (housing, power supplies, cooling, risers, cabling). See performance description for configuration. The subject of Lot 2 is a rack network storage system for internal data storage (raw data of the knowledge base, model artifacts) and as a target for server backups, as well as for the internal operation of a Nextcloud instance for file storage and file exchange within the university network, including associated backups. The system will be operated exclusively on the internal network. The subject of Lot 3 is a compact rack server for the publicly accessible web services of the project (project website, interactive web platform, demonstrator access). A web server with reverse proxy (Apache) will be operated on the system, which receives requests from the internet and forwards them to the internal services. The server is the only externally accessible system and will be operated in its own network segment (DMZ); it does not hold any sensitive user data itself.