GPU Server for AI Model Training and Inference

Delivery of a GPU server: for inference and training of AI models such as Vision-Language-Action and custom AI models for specific robotics applications. The hardware must be 100% natively compatible with NVIDIA Isaac Sim and the underlying NVIDIA Omniverse Framework, ex-works and without any modifications to hardware, firmware, or common software solutions. Furthermore, the hardware must be able to fully execute all common open-source AI models and frameworks based on an NVIDIA CUDA architecture with a Compute Capability of at least 10.3 or higher; offers requiring software workarounds or proprietary adaptations are excluded.

CONTRACTING AUTHORITY

CLASSIFICATION

ADDITIONAL DETAILS