Introduction
Platform Introduction
Last updated
Platform Introduction
Last updated
PrimeHub is a Kubernetes-based platform designed for groups of data scientists. It aims to be an all-in-one, enterprise-level, machine learning platform that provides a seamless MLOps experience.
PrimeHub adopts a group-centric design, which accelerates collaborative development. In PrimeHub, scientists can share datasets, artifacts, and seamlessly collaborate on projects in groups. Models can be developed, deployed, and monitored with full control in a transparent group environment.
Using PrimeHub’s granular controls, administrators can configure access control settings, manage resources, and adjust resource quotas for groups - All facilitating efficient resource utilization.
With the addition of 3rd-party app integration, PrimeHub’s capabilities can be augmented on demand, accelerating the machine learning workflow. These 3rd-party apps also benefit from the access control and resource configurability that is available through PrimeHub, making for a true, all-in-one, MLOps solution.
Cluster Computing
One-Click Research Environments
Easy Dataset Loading
Management of Resource & Quota Privileges
Custom Deep Learning Environments
Enterprise-Class Account Management
Capability Augmentation via 3rd-party Apps
PrimeHub is composed of a User Portal and an Admin Portal. The User Portal provides features aimed at data scientists and team members, and the Admin Portal provides configuration options relevant to platform administrators.
The User Portal provides access to a full MLOps feature set. Data scientists can turn workflows into automated pipelines via Jobs/Recurring Jobs; prepare data and develop trained models from Notebooks; deploy container-wrapped models as services via Model Deployment, and more.
User features
Group admin features (Viewable only to the group admin)
Group switch
Group context
The Admin Portal enables platform administrators to manage all aspects of PrimeHub, such as access-control, resources and quota control, building custom environments via image builder etc.
Platform administrator features
Feature context
Admin Portal entry (Viewable only to the platform admin)