This document is a two-part white paper that covers (a) the benefits of using Comet, a machine learning experimentation platform, in tandem with a Dell EMC AI-enabled infrastructure and (b) technical installation instructions to run Comet directly on a Dell EMC-enabled Kubernetes cluster or as part of a Kubeflow deployment or simple Jupyter notebook server for readers who may want to experiment with this on their own.
Sections:
1. User Story
The data science team user story is a narrative guide through a typical data science team’s workflow, showcasing how the adoption of Comet can dramatically improve a data science team’s ability to work collaboratively, efficiently and transparently, building better models faster. Dell EMC and Comet have decided to make the user story more personal by using character names and roles.
2. Technical Installation Instructions
Dell Technologies provides an extensive portfolio of technologies — spanning workstations, servers, networking, storage, software, and services — to create AI and data analytics solutions that drive successful machine learning and deep learning implementations. Teams can leverage Dell compute infrastructure to build a data science platform and environment using their choice of tools from the AI ecosystem and leverage Comet to make their data scientists more productive. They also have the option of leveraging the Dell EMC portfolio of Ready Solutions and Reference Architectures for AI, which can be utilized with Comet to improve the productivity of your organization.
Comet Enterprise can be deployed on-premises and supports multiple versions of installation from a single machine to distributed microservices. This is an option for organizations that want to host Comet services on their infrastructure for security or privacy reasons and still take advantage of Comet features and insights to improve productivity, collaboration, and visibility across data science teams.
For more information on engineering-validated designs for your applications using industry-leading servers, software, networking and storage visit:
