AI infrastructure
AstroPulse builds AI infrastructure in your own cloud account from features you combine: GPU clusters, add-ons that serve and scale models, access policies, and cost and GPU metrics. Each use case below is a blueprint: which features to combine, in what order, and why. Every feature has its own page with its steps and settings.
Use cases
| Use case | What you build |
|---|---|
| Serve models behind an API | Open models on your own GPUs, as an OpenAI-compatible API on your domain, for callers holding your API keys. Replicas and GPU nodes grow with demand. |
Building blocks every use case shares
| Building block | What it does | Page |
|---|---|---|
| A GPU cluster | GPU nodes in your cloud account, fixed or launched on demand | Create a cluster, GPU node groups |
| Node Autoscaling | Launches a GPU node for pending work and removes empty ones | Node Autoscaling |
| Model Serving | Runs models from Hugging Face or your own storage | Model Serving, Serve a model |
| Workload Queueing | A GPU quota per team | Workload Queueing |
| Gateway, Certificates, DNS Automation | Serve models on your domain over TLS | Expose a model |
| Access policies | Decide who may call a model | Access policies |
| Cost Visibility, Accelerator Metrics | Spend and GPU use per model and team | Cost and metrics |
| Nova | Investigates a failing model or cluster | Nova |