Add End-to-End Workflow Tutorial

[kramaranya]

What you would like to be added?

Create a tutorial in https://sdk.kubeflow.org/en/latest/ showing how all SDK clients work together in a complete ML workflow:

• Train a model with TrainerClient
• Optimize hyperparameters with OptimizerClient
• Register the best model with ModelRegistryClient

Why is this needed?

Currently, each client is documented in isolation. Users don't see how the pieces fit together or understand the intended workflow. A unified tutorial demonstrates the value of the SDK as a cohesive platform rather than separate tools.

Love this feature?

Give it a 👍 We prioritize the features with most 👍

[kramaranya]

/remove-label lifecycle/needs-triage
/good-first-issue
/area docs

[google-oss-prow]

@kramaranya:
This request has been marked as suitable for new contributors.

Please ensure the request meets the requirements listed here.

If this request no longer meets these requirements, the label can be removed
by commenting with the /remove-good-first-issue command.

Details

In response to this:

/remove-label lifecycle/needs-triage
/good-first-issue
/area docs

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[farhann-saleem]

I'd like to work on this!

My approach:

1. Create a simple end-to-end example (using MNIST or a tabular dataset)
2. Show the complete workflow: TrainerClient → OptimizerClient → ModelRegistryClient
3. Focus on demonstrating how the SDK clients integrate, with clear code examples and explanations for each step
4. Follow the existing tutorial structure and style in the SDK documentation

I'll ensure all code snippets are clear, well-commented, and the tutorial provides practical value by showing how these components work together in a real ML workflow.

Currently reviewing the existing documentation to match the style and format.

[andreyvelich]

Looks great! Also, let's ensure we run this example as part of our E2Es: https://github.com/kubeflow/sdk/blob/main/.github/workflows/test-e2e.yaml
So it is functional.

/area examples

[farhann-saleem]

Thanks @andreyvelich! I appreciate the feedback.

I'll refactor the tutorial to be E2E-test-ready. I've reviewed the existing E2E workflow and see the pattern with pytest tests in tests/e2e/.

One clarification: For the Model Registry component, I notice the E2E tests use test clusters - should I mock the Model Registry connection, or is there a test instance I can reference?

I'll update the PR to include:

• Functional example code in examples/
• Corresponding E2E test in tests/e2e/
• Documentation that references the working example

This might take some time to implement properly, but I want to ensure it meets your testing standards.

[asmaa34937]

Hi @andreyvelich @kramaranya , I'm a CS senior student specializing in AI, and I’ve been following the discussion here.

I’m currently working on a Multimodal AI graduation project that involves complex ML pipelines (from preprocessing to model evaluation), so I have a solid understanding of how Trainer, Optimizer, and Registry clients should integrate in a real-world E2E workflow.

I see @farhann-saleem has proposed a great approach. Since this is a high-priority feature, I’d love to collaborate or take over if more hands are needed to ensure the example is functional and meets the E2E testing standards mentioned. I'm ready to start working on the code structure and documentation immediately. Looking forward to your guidance!