Request to be User Group Contributor for Kubeflow 1.5 for the AutoML Working Group

[d-gol]

Hello everyone, thank you for the great work on Kubeflow.

My name is Dejan Golubovic, I am a software engineer at CERN, where we run Kubeflow in production.

My experience with Kubeflow includes:

• Deploying a cluster on private cloud, using custom ArgoCD repository
• Customizing Kubeflow components to fit our use cases
  • Maintaining a fork of Kubeflow repo
  • Custom centraldashboard, jupyter-web-app, kale and kfp setup
  • Adding envoyfilters, roles and rolebings when necessary
  • Integrating CERN network file system EOS and Kerberos authentication
• Maintaining the private cloud instance, fixing ongoing issues
• Providing examples and documentation for CERN users
• On-boarding users and helping run workloads (pipelines, Katib, TFJobs, PytorchJobs...)
• Running scalable model training on a GCP and Azure instances (up to 128 GPUs)

Our previous Kubecon talks can be seen here:

• Kubecon 2021 NA - https://www.youtube.com/watch?v=sBbfUPaTCVA&ab_channel=CNCF%5BCloudNativeComputingFoundation%5D
• Kubecon 2021 Europe - https://www.youtube.com/watch?v=HuWt1N8NFzU&ab_channel=CNCF%5BCloudNativeComputingFoundation%5D

While running the production instance, we are gathering feedback from CERN users.
I would like to help implement some of the features for which we see a need.

One such feature is showing Katib logs on the UI: #971
I could help with this in the following weeks.

Informing WG leaders - @kubeflow/wg-automl-leads

Please let me know if you need more information.

[andreyvelich]

Welcome Dejan and thank you for your interest!
That would be great if you could help us with that issue.

I suggest join our next AutoML Community Meeting on December 1st to discuss the implementation details.

cc @johnugeorge @gaocegege

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.