Wind Tunnels for ASI - Building Infrastructure for Advanced AI Systems Testing

Learn to build and deploy private AI infrastructure using open-source models in this 17-minute conference talk from the Open Compute Project. Discover how to create bespoke, perimeter-contained AI systems using models like DeepSeek on high-performance infrastructure including HPE with NVIDIA GPUs. Explore the development of tools and scripts for deploying current state-of-the-art open-source models as APIs that can be consumed by custom interfaces such as coding bots. Understand the broader vision of building scaffolding for testing and developing future AI approaches beyond traditional LLMs-LRMs, focusing on frontier AI reasoning systems and guided program synthesis. Gain insights into the proof-of-concept bootstrap process that enables rapid deployment of functioning AI instances, with all tools and techniques made available through the OCP AI repository on GitHub.