This Week in CN AI (20-25 May 2024)

This week, the Cloud Native AI (CNAI) Working Group has made progress in several key areas. The whitepaper on the challenges of scheduling cloud-native AI workloads has been refined and is proposed to be titled “Challenges of Scheduling Cloud Native AI Workloads Whitepaper.” Additionally, a presentation on sustainable AI practices was delivered, focusing on optimizing power consumption for popular systems and introducing new tools and metrics. The group continues to seek community contributions and feedback to enhance these initiatives and drive forward the development of efficient, scalable AI operations in cloud-native environments.

Key Developments

  • The working group had its weekly meeting, which you can watch here.
  • Challenges of Scheduling Cloud Native AI Workloads Whitepaper: The group is doing another round of review with goals to publish soon. The issue is being tracked here.
  • Sustainable AI Presentation: Marlow Warnicke, Chen Wang, and Thomas Parnell presented on Cloud Native Sustainable AI, discussing power consumption of popular systems like Bitcoin and ChatGPT, and the scope of including observability, analytics, and reduction tools.
  • Kubeflow Intro Blog Review: The Kubeflow intro blog has gone through several rounds of review. The goal is finalize the blog and publish towards the end of week,
  • Sustainable AI Initiative: Discussions on power metrics, tools like Kepler (in sandbox), and fmperf (IBM might donate) were highlighted. The next steps involve collaboration with the sustainability tag via Slack.
  • Project Dashboard: Overview on the project dashboard and where the group tracks its issues.

Events and Participation

Check upcoming events here.

How to Get Involved

  • Sustainable AI Initiative: Join discussions and contribute to the development of sustainable AI metrics and tools by participating in the sustainability tag channel on Slack #tag-env-sustainabiltiy-best-practices.
  • Kubeflow Blog Review: Participate in the review process by providing feedback on the Kubeflow blog.

The CNAI Working Group welcomes participation! Here’s how you can contribute:

  • Project Dashboard Contributions: Contribute to CNAI by submitting issues/PRs to your favorite initiative, the issues will show up on the project dashboard. Examples and guidelines can be found on GitHub.