The hybrid approach leave us the management of the infrastructure that does not touch your data. In this way you don't need to ship your data outside your infrastructure.
The Rockfish Workbench is made of various component for what concerns the hybrid deploy Rockfish will provide to you a control plane made of a web ui and an HTTP API, you will have to deploy what we call the data plane.
In practice one or many Python workers that will register to our API and will run training and generation workflows and a Redis instance that workers use to exchange data in between jobs.
In this way you are under control of your data, we'll take care of workflow monitoring and lifecycle.
Cuttlefish
Cuttlefish is a Python application that registers to the control plane provided by Rockfish and executes various actions here a short list to have an idea:
- upload and download data from and to various objects stores such as: GCP, S3, MinIO
- Train and generate data with various algorithms
- Amplify your generated data to continuously an quickly generate more
Self hosted worker
If you have a machine learning team or you want to train your own way our platform support self hosted worker that you write and hook to our API to extend the capabilities we provide with your specific needs.
Deploy
Check our deploy page to know how we can help you releasing Rockfish Workbench.