CoreWeave targets AI inference with full-stack optimization

CoreWeave is building managed services for training, post-training and inference atop its GPU infrastructure to address AI inference bottlenecks. VP Urvashi Chowdhary highlighted CoreWeave RL Rollouts, built on Nvidia's Dynamo framework, which improved model reload latency by 15x during reinforcement learning workflows. These capabilities are now part of CoreWeave Forge, a platform connecting serving, observability, post-training and evaluation, free to start with paid tiers available. A theCUBE Research survey found one healthcare customer's inference workload share grew from 10% to 40% year over year.