Velda is a serverless GPU platform that lets developers run distributed AI training and batch inference jobs directly from their local environment by simply prefixing commands with vrun . It eliminates the need for Docker images, Kubernetes manifests, and dependency management, providing instant access to cloud or cluster GPUs with a familiar local development experience. The platform includes a managed cloud option with VS Code browser access and prebuilt templates, plus an enterprise tier for self-hosted or dedicated infrastructure.
Funding
Funding not disclosed
Founders
Product
Problem
Developers running AI workloads face significant operational overhead when moving from local development to cloud or cluster environments. They must build container images, write Kubernetes manifests, manage dependencies, and navigate complex infrastructure setup, which slows iteration and creates friction between development and production.
Solution
Velda provides a serverless GPU platform that lets developers run distributed AI training and batch inference jobs directly from their local environment by prefixing any command with `vrun`. The platform eliminates the need for Docker images, Kubernetes manifests, and dependency drift, giving users instant access to cloud or cluster GPUs while preserving their local development experience. Developers can launch VS Code in their browser or connect their favorite IDE, starting with prebuilt templates for frameworks like PyTorch and then customizing their workflow. The platform scales seamlessly from individual experimentation to production-scale training, serving, and data pipelines, all through the same simple command interface.
Target Audience
Velda targets AI/ML developers and data scientists who need GPU compute for training and inference workloads, from individual practitioners and small teams using the managed cloud to large organizations requiring self-hosted or dedicated infrastructure.
Features
- `vrun` command prefix that runs any command on cloud or cluster GPUs without image builds or manifests
- Instant VS Code browser access or IDE connectivity for a familiar development environment
- Prebuilt templates for PyTorch and other ML frameworks with base dependencies pre-configured
- Support for distributed training, batch inference, serving, and data pipeline workloads
- Managed cloud option with free monthly credit for individuals and small teams
- Enterprise tier with self-hosted or dedicated infrastructure and premium support