Community Liaison

NVIDIA and CoreWeave deploy production AI infrastructure for agentic workloads. At the CoreWeave Fully Connected conference in San Francisco this week, NVIDIA and CoreWeave unveiled production-ready infrastructure designed to bridge the gap between AI model training and agentic deployment. The announcement centers on the immediate availability of NVIDIA Vera Rubin NVL72 systems paired with Spectrum-X 102.4T Ethernet networking on the CoreWeave Cloud platform. Cognition, the developer of the Devin AI software engineer, has been selected as the inaugural customer to run production inference workloads on the new architecture. Early benchmarking demonstrates that Vera Rubin delivers up to a 4.8x increase in total token throughput for complex SWE-2 coding tasks compared to the preceding GB200 NVL72 baseline, significantly accelerating real-time code generation and multistep reasoning. To support the computational demands of agentic AI, CoreWeave also introduced NVIDIA Vera, a CPU specifically engineered for high-concurrency agent environments. By housing 128 processors and 11,264 cores per rack, the system supports over 11,000 concurrent isolated agent environments. Integrated with CoreWeave Sandboxes and managed through Spectrum-X switches and BlueField-4 DPUs, the Vera CPU achieved more than a 3x improvement in sandbox startup times and a 1.7x performance gain on the Terminal-Bench evaluation suite, ensuring low-latency, secure communication across distributed agent workloads. Complementing the hardware rollout is CoreWeave Forge, a unified development environment designed to close the traditional divide between production telemetry and model refinement. Forge consolidates Weights & Biases, OpenPipe post-training capabilities, and the marimo notebook framework into a single workflow. The platform introduces CoreWeave ARIA for automated experiment analysis and code iteration, Agent Lens for production observability that reduces issue resolution costs by 50%, and serverless reinforcement learning capabilities that accelerate training by 1.4x while cutting costs by 40%. A newly private-previewed feature, RL Rollouts, leverages NVIDIA Dynamo to load model checkpoints into live deployments without service interruption, enabling continuous reinforcement learning alongside production inference. Enterprise adoption of the NVIDIA-CoreWeave stack is accelerating rapidly. Beyond Cognition, initial Forge users include Canva, Capital One, and MasterClass. In the healthcare sector, Ennoble Care has selected CoreWeave to deploy reserved NVIDIA RTX PRO 6000 GPU capacity for scaling clinical documentation agents across Medicare patient networks. The partnership maintains CoreWeave's position as the exclusive Platinum-ranked provider across SemiAnalysis ClusterMAX benchmarks and continues to serve nine of the top ten AI laboratories. By co-engineering infrastructure that scales from prototype deployment to continuous production feedback, NVIDIA and CoreWeave are establishing a standardized pipeline for transforming experimental AI agents into reliable, large-scale enterprise systems.

Ennoble Care expands national reach by entering five new communities. Published: September 17, 2026 ARLINGTON, VA - September 17, 2026. Ennoble Care today announced its expansion into five new communities - in Florida, Kentucky, Ohio, Mississippi, and Texas - bringing home-based primary, palliative, and hospice care to more medically frail and homebound patients. An expanded care coordinator team and Ennoble Care's proprietary electronic health record platform support the growth, which continues through a combination of organic expansion, value-based partnerships, and thoughtful additions of high-quality house call practices.