cnvrg-operator Certified

Compose a pipeline from datasets through data preparation to many training models running in parallel, with the best performing model picked for deployment. Models in deployments are monitored and can trigger a rerun of the pipeline. Each task in the pipeline can run on different resource (GPU or CPU) and in different Kubernetes cluster (on-prem or cloud).

Onboard your company chosen models and share across researches, data scientists, DevOps and MLOps engineers. The models are ready to run and attach to any pipeline with a single click. Any AI library model can be tuned based on hyper parameters.

Onboard machines (GPUs and CPUs) and cloud instances, and create 'slices' of compute - compute templates (such as 2x V100, or 8 CPU cores) to use across the teams. Attach compute template of your choice to any trained model, accomplishing multiple experiments running in parallel, utilizing the resources in the most optimized way.