Cast AI Application Performance Automation Platform

Cast AI was founded in 2019 by Yuri Frayman, Leon Kuperman and Laurent Gil and is headquartered in Miami, Florida, with most development in Lithuania, Poland, Romania and Bulgaria. It addresses overprovisioned Kubernetes clusters and scarce, expensive GPU capacity that inflate cloud bills for engineering teams.

The platform connects to EKS, GKE, AKS, OpenShift and on-premise clusters, starts in read-only mode and then automates rightsizing of workloads, node consolidation, spot instance use and autoscaling. Additional modules include Storage Autoscaler, Database Optimizer for PostgreSQL and MySQL, GPU sharing via time-slicing and MIG, OMNI Compute for sourcing GPUs across clouds and regions, plus OpsPilot, agentic workflows and an MCP server. The site says it is trusted by 2,100+ companies, including Akamai, BMW Group, FICO, NielsenIQ and Swisscom.

• Workload rightsizing: Continuously adjusts CPU and memory requests of Kubernetes workloads.
• Cluster autoscaling: Consolidates nodes and selects cheaper instances including spot, working with Karpenter or its own autoscaler.
• GPU optimization: Shares GPUs through time-slicing and MIG and sources GPUs across clouds with OMNI Compute.
• Storage and database optimization: Expands volumes before they fill and caches repeated database reads.
• Cast Radar: Reports cost, error rate and latency across clusters.
• AI operations: OpsPilot, agentic workflows and an MCP server answer questions and open pull requests for fixes.

A Vietnamese e-commerce or fintech company running EKS or GKE can automatically rightsize clusters and use spot capacity to cut monthly cloud bills. An AI startup training or serving models can share GPUs and source extra GPUs across clouds during shortages. A bank's platform team can monitor cost and latency across multiple clusters from one view.

Customer stories on the site report 50-80% lower Kubernetes costs at Banking Circle and 65% lower production-cluster costs at Jisr without hurting availability. Automation frees engineering time, with Altruist citing 85-130 hours recovered per month. GPU sharing and cross-cloud GPU sourcing help AI teams obtain capacity when local GPUs run out. A 30-day free trial with no credit card lets buyers measure savings before committing.

Kubernetes has become the default runtime for digital products, while AI workloads have pushed GPU spending and shortages to the top of infrastructure budgets. FinOps practice is shifting from dashboards to automated remediation, which suits Southeast Asian digital businesses running on AWS, Google Cloud and Azure regions in Singapore, Jakarta and Bangkok.

Founded in 2019 and headquartered in Miami. Raised a USD 108 million Series C in April 2025 co-led by G2 Venture Partners and SoftBank Vision Fund 2; TechCrunch sources put the valuation near USD 900 million. Partner status: AWS Qualified Software, Google Cloud Premier Partner, CNCF and FinOps Foundation member.