Cast AI

Cast AI builds software for cloud-native companies and platform engineering teams to automate Kubernetes operations, manage costs, and improve container security. Its Kubernetes platform combines cost reporting with infrastructure changes: it scales clusters, packs workloads onto nodes, and adjusts container resources based on actual usage. It supports AWS, Google Cloud, and Azure, with workload optimization and resource consolidation available for on-premises and other environments through Cast AI Anywhere. OpsPilot, its AI agent for DevOps and SRE teams, uses application telemetry—including resource use, latency, and stability signals—to create and adjust workload optimization policies.

Co-founders Yuri Frayman, Leon Kuperman, and Laurent Gil developed Cast AI after their experience with rising cloud bills at their previous startup, Zenedge, which Oracle acquired in 2018. That infrastructure-efficiency focus now extends to GPU capacity. OMNI Compute, introduced in January 2026, brings capacity from other providers and regions into existing Kubernetes clusters as native compute, without application code changes. Teams running AI inference can access external GPUs while retaining control over where workloads execute.

Cast AI also develops Kimchi Coding, an open-source, terminal-native coding agent that reached general availability in July 2026. Its harness routes work according to task complexity and model cost, using frontier models for harder steps and open-weight models for the rest. Feedback loops score generated code, while Cast AI’s infrastructure optimization supports its self-hosted inference. For enterprise development teams, Kimchi adds spend limits and cost reporting by developer, team, and project. It can run on Cast AI’s inference cloud or inside a customer’s own environment, including air-gapped deployments.