CoreWeave Launches Forge Platform to Connect AI Improvement Loop and Expand Full-Stack AI Cloud

Forge moves through five stages: run, observe, curate, improve, evaluate and repeat. CoreWeave added several capabilities targeting the hardest parts of that cycle. Agent Lens lets teams inspect what their agents are doing, and Registry models checkpoints and agent configurations. RL Rollouts and model distillation handle improvement, while CoreWeave's AI Research and Iteration Agent, or ARIA, is now generally available to help users spot patterns. "It's our new way of thinking about the AI loop in a way that allows you to go from first agent to your very best agent as quickly as possible," said Susanne Seitinger, vice president of product marketing at CoreWeave, in an interview on theCUBE. "The faster you learn, the faster you ship, the faster you get value."

CoreWeave was the first to bring up and validate Nvidia Corp.'s Vera Rubin platform, and the company recently earned SemiAnalysis' Platinum ClusterMAX rating for a third consecutive time. Jean English, CoreWeave's chief marketing officer, said the AI era calls for building from first principles. "It's so much beyond the GPU," she said. "It's about a full-stack AI cloud. That means that the tooling that we need, there's got to be other partners that we work with." The broader platform brings together compute, networking, storage, software and operations to support AI workloads in production. Customers can begin with one use case, and many are currently looking at inference, according to English. "The model makers are definitely training big frontier models," she said. "The enterprises are working on the best use case to show results and impact."

The ecosystem is the other half of the plan. CoreWeave also introduced a partner network built on tested, co-engineered integrations. "It's really about giving folks more recipes and playbooks and use cases and solutions so they can get to value faster," Seitinger said. She added that the timing matters because the workload mix is changing fast. Cognition AI Inc. is already running production workloads on CoreWeave's first Nvidia Vera Rubin NVL72 racks.

CoreWeave built its platform specifically for AI workloads instead of adapting infrastructure designed for an earlier generation of web applications. Clients arriving from other clouds often cite gaps in speed, performance, capacity or tooling, English explained. "In this market, I think the AI era calls for building from first principles," she said. "That's why we are purpose-built. It was really built from the ground up … and I think that's why, when clients are with other clouds, they're not able to move as fast and able to get the performance they need."