CoreWeave Launches Forge Platform to Accelerate Agentic AI Improvement Cycles
CoreWeave introduces Forge, a full-stack platform designed to unify workflows for running, observing and continuously improving AI agents in production environments.
Key points
- CoreWeave Forge provides a shared workflow for business units and machine learning teams to manage the continuous AI improvement loop.
- The platform features Agent Lens for inspection, Registry for configuration management, and ARIA for pattern recognition during model iteration.
- Cognition AI is already running production workloads on CoreWeave infrastructure, including new Nvidia Vera Rubin NVL72 racks.
CoreWeave Inc. has launched Forge, a new full-stack platform designed to accelerate the continuous improvement of agentic AI systems. According to Susanne Seitinger, vice president of product marketing at CoreWeave, the offering aims to unify workflows for business units and machine learning teams. This integration allows organizations to move from initial agent deployment to optimized performance more rapidly by connecting disparate tools and processes.
The platform addresses the reality that enterprises racing to deploy agents find the improvement process never truly ends. Agents must be run, observed, evaluated and improved continuously. Forge structures this cycle into stages: run, observe, curate, improve, evaluate and repeat. Each pass through the loop is intended to build upon the previous one, with the speed of learning serving as the primary metric for success, according to Seitinger.
Platform Capabilities and Workflow
Several new capabilities within Forge target the most difficult aspects of the AI development cycle. Agent Lens allows teams to inspect what their agents are doing, while Registry manages checkpoints and agent configurations. For the improvement phase, the platform supports RL Rollouts and model distillation. Additionally, CoreWeave’s AI Research and Iteration Agent, known as ARIA, is now generally available to help users identify patterns.
Seitinger explained that distilling knowledge from large models into specific use cases is critical for enterprises. The platform is designed to facilitate this translation, ensuring that general model capabilities are effectively applied to organizational needs. This approach supports the transition from experimental models to production-ready agents that can operate within complex business environments.
Market Context and Infrastructure
The timing of Forge’s launch coincides with rapid changes in AI workload demands. Cognition AI Inc. is already running production workloads on CoreWeave’s infrastructure, specifically utilizing the company’s first Nvidia Corp. Vera Rubin NVL72 racks. This deployment highlights the growing reliance on specialized hardware to support intensive AI training and inference tasks.
Seitinger noted that post-training and continuous improvement are a reality for organizations deploying AI. The ability to iterate quickly on models after initial deployment is becoming a competitive advantage. CoreWeave’s platform aims to provide the necessary infrastructure and workflow tools to support this ongoing cycle of refinement, ensuring that AI systems remain effective as requirements evolve.
What should you do?
Organizations deploying AI agents should assess their current workflows for bottlenecks in observation and evaluation. Teams should verify that their infrastructure supports continuous integration and deployment of model updates. Monitoring agent performance in production is essential to identify areas for improvement.
What to do and how to stay safe: CoreWeave
- Audit current AI agent workflows to identify gaps in observation and evaluation processes.
- Ensure infrastructure can support continuous model updates and version control for agent configurations.
- Implement robust logging and inspection tools to monitor agent behavior in production environments.
- Plan for regular model distillation and refinement cycles to maintain performance over time.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the primary goal of CoreWeave Forge?
CoreWeave Forge aims to speed up the AI improvement loop by providing a shared workflow for running, observing and improving agents.
What tools are included in the Forge platform?
The platform includes Agent Lens for inspection, Registry for configuration, RL Rollouts for improvement and ARIA for pattern recognition.
Which company is already using CoreWeave’s new infrastructure?
Cognition AI Inc. is running production workloads on CoreWeave’s Nvidia Vera Rubin NVL72 racks.
