Aranya
Category: AI Infrastructure
Aranya builds ClusterdOS, an open-source, GitOps-native cluster-scale operating system that turns Kubernetes into a self-healing platform for AI inference infrastructure. Aranya was founded in 2025. The company is led by Christian Bhatia Ondaatje. Based in San Francisco, California, United States. Team size: 11-50. Total funding raised: $1.2M. Latest round: Pre-Seed. Key investors include Seedcamp; Asylum Ventures; Founder Collective.
- Founded
- 2025
- Headquarters
- San Francisco, California, United States
- Team size
- 11-50
- Total funding
- $1.2M
Value proposition
ClusterdOS eliminates the need for dedicated platform teams by providing a self-healing Kubernetes-based operating system that handles the full cluster lifecycle — bootstrapping, maintaining, and upgrading — with a single Git push, reducing cluster setup time from weeks to under 48 hours and cutting downtime by up to 90%.
Products and solutions
ClusterdOS (flagship open-source cluster operating system — GitOps-native, self-healing Kubernetes distribution); Vibecluster (upcoming team-level inference scaling management tool); Enterprise Support (24/7 monitoring, security patching, compliance, custom integrations, managed Kubernetes services)
Unique value
ClusterdOS is the first "personal cluster operating system" that makes Kubernetes-grade orchestration accessible without a dedicated platform team — define your entire cluster in Git, toggle applications like feature flags, and get production-ready defaults with self-healing capabilities. Provider-agnostic (any cloud, bare metal, or GPU compute).
Target customer
AI inference companies, GPU cloud providers (neoclouds), bare-metal AI infrastructure providers, and engineering teams running compute-intensive GPU workloads at scale who need production-grade Kubernetes without dedicated platform engineering teams.
Industries served
AI Infrastructure, GPU Cloud Computing, AI Inference, Distributed Computing
Technology advantage
GitOps-native architecture with ArgoCD; self-healing Kubernetes that auto-recovers from failures; single-file cluster definition; provider-agnostic (works across AWS, GCP, bare metal); open-source (GPL v3); production-tested across 1,700+ GPUs with enterprise partners; reduces cluster setup from 2-6 weeks to under 48 hours; 90% reduction in cluster downtime.
How they differentiate
Unlike hyperscaler managed K8s (EKS/GKE/AKS) which offer barebones Kubernetes requiring significant platform engineering, ClusterdOS is a full operating system layer on top of Kubernetes that is self-healing, GitOps-native, and provider-agnostic. Unlike GPU cloud competitors (RunPod, Modal, Baseten) which offer managed inference platforms, Aranya gives teams ownership of their clusters with an open-source OS they control, while eliminating the need for dedicated platform teams.
Main competitors
RunPod; Modal; Baseten; Together AI; Nebius; Google GKE/AWS EKS/Azure AKS (hyperscaler managed K8s)
Key partnerships
Hydra Host (NVIDIA cloud partner, bare-metal AI deployment — first major partnership deploying ClusterdOS across 1,700+ GPUs); Baseten (AI inference company, testimonial from Head of Global Capacity Ed Shrager)
Notable customers
Hydra Host (NVIDIA cloud partner); Baseten (AI inference company)
Major milestones
April 2026: Emerged from stealth with ClusterdOS launch and Hydra Host partnership; May 2024: Initial ClusterdOS repository created on GitLab; 2025: Company founded; Raised $1.2M pre-seed from Seedcamp, Asylum Ventures, and Founder Collective
Growth metrics
Deployed across 1,700+ GPUs at launch (April 2026); 90% reduction in cluster downtime for Hydra Host; cluster setup time reduced from 2-6 weeks to under 48 hours
Market positioning
Early-stage startup positioning at the intersection of AI inference infrastructure and Kubernetes orchestration. Aranya targets the "personal cluster" paradigm shift — making distributed computing ownership accessible to individual teams, analogous to how personal computers democratized mainframe computing. Strategically positioned as inference becomes the dominant GPU workload (the "new mining").
Geographic focus
United States (San Francisco-based), with global reach through open-source distribution and cloud-agnostic architecture
About Christian Bhatia Ondaatje
Ex-Founding Platform Engineer at Hyperbolic; Ex-3rd Engineer (Federated Kubernetes Engineer) at Crusoe Energy Systems; B.A. in Computer Science from Harvard University
Official website: https://aranya.tech/