Case Study: Center for AI Safety accelerates AI research and cuts storage costs with Weka

How the Center for AI Safety empowers research into safe and responsible AI through the CAIS compute cluster

The Center for AI Safety (CAIS), a San Francisco-based nonprofit focused on advancing safe and responsible AI research, needed a better way to support growing demand for its GPU-based compute cluster. As more researchers began running LLM training and AI safety experiments, CAIS faced slow model training times, storage performance limits, underutilized infrastructure, and rising costs from data copying and storage sprawl.

To address these challenges, CAIS implemented the WEKA Data Platform in Converged Mode. With WEKA, CAIS reduced storage costs by 90%, improved storage performance by 5x, and increased resource utilization from less than 20% to 100%. The improved infrastructure helped CAIS grow its active research community from 30 to more than 200 users while supporting more projects at much lower cost.

Center for AI Safety

Stephen Basart

R&D Lead