Share Your Experience & Suggestions for Kubeflow SDK

[andreyvelich]

Hi Kubeflow Community! 👋

We’re shaping the next phase of the Kubeflow SDK roadmap, and we’d love your input to ensure we focus on what matters most to you.

Please use this thread to share:

⚙️ Challenges you face when running AI workloads at scale or using AI platforms in production
💭 Ideas or feature requests for improving the Kubeflow SDK
🧩 Feedback on existing APIs, such as TrainerClient(), OptimizerClient(), or local backends (Container + Subprocess).

Your insights will help us identify key priorities and refine how the Kubeflow SDK supports running distributed AI workloads at scale.

For more structured feedback, please also complete this short survey: (TODO: Add link)

Thank you for helping us build the future of Kubeflow SDK together! 🚀

For more details about the Kubeflow SDK, check out:

• The announcement blog post
• Kubeflow Trainer documentation
• Tentative Kubeflow SDK ROADMAP