Pipeline Core

An overview of pcore

It takes skilled DevOps teams months to setup reliable production systems that run GPU or CPU compute for AI. Pipeline Core (pcore) is a set of tools to automate the deployment and management of production grade AI infra structure in your cloud or on-prem.

There are two components to pcore: it's serverless GPU/CPU compute engine, and the command line interface (CLI) thats used to deploy or manage a pcore cluster. There are an array of features of pcore that make running AI in your use case much easier:

  1. Multi-model/multi-environment GPU sharing
  2. Autoscaling or fixed number of GPUs/CPUs
  3. Business level control (optimise routing and scaling for cost or speed)
  4. Cold start reduction via preemptive caching
  5. Model versioning
  6. CI/CD integrations
  7. Deploy on a cloud provider or on-prem
  8. CLI (pcore)
  9. Auto-generated dashboards and monitoring tools

Note: PipelineCore is currently in Beta with all features from former API versions being brought forward. You can see this feature parity progress in the Features & roadmap page.