Capabilities
Serverless Pods provide developers and data scientists with instant access to powerful compute resources (CPU/GPU) without managing long-lived infrastructure. Each pod is an isolated Ubuntu-based container bundled with compute, networking, and optional storage.
Pods are optimized for ephemeral, event-driven workloads like inferencing, API services, and batch jobs. Resources are only consumed when the pod is running, helping teams reduce idle costs while retaining flexibility.
Key Capabilities¶
- Flexible container environments supporting any container image, including GPU-accelerated ML images with pre-installed CUDA and PyTorch versions
- Hardware resources: vCPU, memory, and GPUs (optional)
- Container volume for OS and temporary storage
- Network connectivity via SSH
- Optional persistent disk for stateful workloads
Why Use Serverless Pods¶
-
Instant & Flexible Environment
Quickly deploy compute-intensive tasks like ML training, data processing, or rendering. -
Transparent Control
Full customization of OS, software stack, networking, and storage. -
Cost-Effective & Scalable
Scale pods down to zero when not in use and pay only for active resources.