Skip to main content

Platform Onboarding

Welcome! This guide explains the basic concepts, prerequisites, and steps to get started with Connected AI Platform (CAIP) and your Managed Kubeflow Workspace.

📖 Glossary / Nomenclature​

Managed Kubeflow Workspace​

A Managed Kubeflow Workspace is your production-ready Kubeflow environment, hosted in either the EMEA Hub or the CN Hub, fully operated by the CAIP platform team.

After initial provisioning, the main entry point for Data Science and ML development is the Kubeflow Web UI. Access your workspace dashboard using its namespace name in the URL:

RegionURL Pattern
Rest of the World (excl. China)https://kubeflow.caip.bmw.cloud/?ns=NAMESPACE
Chinahttps://cd4ml.cn-north-1.prod.mcaip-cn.cv.bmw.com.cn/?ns=NAMESPACE

After following this URL, you will be forwarded to BMW's SSO screen (WebEAM) and prompted to enter your BMW credentials. After signing in, ensure the correct namespace is selected under the profile tab.

✅ Prerequisites​

Before you can get started with the Connected AI Platform, you need to ensure you have access to the right tools for managing your code and data.

As a use case owner, you are responsible for managing your own GitHub repositories and organizations to store and version your code. In addition, you must manage your data access through BMW's Cloud Data Hub (CDH), which is the enterprise data platform for storing and sharing data at BMW.

If your use case needs to read data, you must set up a CDH Use Case and create a CDH Environment to access the required data. If your use case needs to provide (upload) data to CDH, you must set up a CDH Provider and a CDH Environment. Each CDH Environment can be linked to your Managed Kubeflow Workspace, enabling seamless and scalable data integration for your projects.

Connected AI Platform provides an easy and scalable integration with CDH for your use cases.

APP-ID​

Each use case on the Connected AI Platform requires an APP-ID. The APP-ID is a unique identifier for your application within BMW IT, used to organize and manage applications in the software portfolio. Having an APP-ID is essential for linking your use case to the correct IT Product and ensuring proper documentation and governance. You can read more about how to create and link an APP-ID in our step-by-step guide Create and Link an APP-ID.

GitHub​

  1. Access to ATC GitHub Please make sure you are able to access ATC GitHub by having the necessary RightNow Roles assigned to your account. Learn more on how to get access to ATC GitHub here. You can find the whole ATC GitHub Documentation here.

  2. Own GitHub Organization
    If you do not yet have a GitHub Org on ATC GitHub, you can request one here: Request GitHub Org. You need to specify a unique organisation key, a display name and organisation admins. You need to specify a link to a BMW IT Product or Application. If you do not have an own Application (APP-ID) or IT-Product (SWP-ID), feel free to use the SWP-ID of Connected AI Platform: SWP-4628.

CDH Use Case or Provider​

  1. Access to Cloud Data Hub (CDH)
    CDH is BMW's enterprise data platform for managing, consuming, and providing data at scale. To access or provide data for your use case, you need to set up the appropriate CDH resources:

  2. CDH Use Case (Data Consumer)
    If your use case needs to read or consume data from CDH, you must create a CDH Use Case and a CDH Environment. Via the environment you can link your Managed Kubeflow Workspace to the required datasets.

  3. CDH Provider (Data Provider)
    If your use case needs to provide or upload new data to CDH, you must set up a CDH Provider and a CDH Environment. After setting up the CDH Link, this setup allows your Managed Kubeflow Workspace to write data into CDH and share it with others.

A CDH Environment can be linked to your Managed Kubeflow Workspace, enabling seamless data integration for both consuming and providing data. Learn more about CDH and how to request access in the CDH Documentation.

Onboarding Introduction​

Once you have fulfilled the prerequisites you can start getting onboarded to Managed Kubeflow!

Once you have a Managed Kubeflow Workspace, you can request additional features associated with it, such as linking a GitHub repository via CI Trust or connecting to a CDH Environment.

The following section will explain how to get started with different CAIP Feature Offerings. The guide outline a Golden Path, a predefined routes in a platform that outline the best practice. If you already know what you are doing, you can defer from this path and use our Service Requests directly.

What are golden paths?

Platforms offer developers quick access to resources and tools to turn concepts into reality. From code libraries to testing frameworks and collaborative spaces, platforms have become accelerators for innovation. Golden or paved paths are predefined routes in a platform that outline the best practices, steps, and required tools to achieve a specific goal. They are well-structured pathways that guide developers through building, deploying, and testing applications.

To understand this better, think of the process of building and deploying a web application. A golden path for this use case involves a development environment for building with node.js, configuring docker containers and Git for version control, setting up Jenkins for CI/CD, Kubernetes for container orchestration, and MongoDB for data. All this, including building, testing, and deployment, is automated.

This golden path can be provided as a template or guided workflow for new developers to help them walk through the golden path, giving them all the necessary tools and configurations to start the development process.

You can read more about golden paths on platformengineering.org.

Onboarding to Managed Kubeflow​

1 - Create Managed Kubeflow Workspace​

To get started, request a Managed Kubeflow Workspace via Service Request. A Managed Kubeflow Workspace is your dedicated deployment of Kubeflow within our managed infrastructure. We will deploy and operate the full Kubeflow Stack for you, so that you can use Kubeflow Notebooks, Kubeflow Pipelines and Kubeflow Endpoints (based on KServe). The Managed Kubeflow Workspace contains a Model Registry (MLflow) as well.

Kubeflow Notebooks & Cloud Development Environment​

A Kubeflow Notebook is a Jupyter Notebook running on CAIP, referred to as a Cloud Development Environment (CDE). Access this service directly through the Web UI under the Notebooks section. After initial configuration, you can start prototyping, exploring data, and training models interactively while CAIP manages infrastructure and data access in the background.

Kubeflow Notebooks are used for:

  • Data exploration
  • Prototyping
  • Model training and evaluation

Kubernetes/EKS Cluster Access​

Normally, Kubeflow abstracts away the need to access the underlying Kubernetes cluster. However, for advanced debugging or resource manipulation, you can access the cluster from a terminal inside the CDE (e.g., VSCode Server Notebook), where the required access rights are already assumed. We suggest using k9s for cluster resource management and debugging.

2 - Link your GitHub Repo​

To enable automated CI/CD workflows and securely manage your code, you need to (1) create a GitHub repository/organization and (2) link it to your Managed Kubeflow Workspace using CI Trust.

Step 1: Create a GitHub Repository/Organization

  1. Ensure you have access to ATC GitHub with the necessary RightNow Roles. Learn more about access here.
  2. If you do not yet have a GitHub Organization, request one here. You will need to specify:
    • A unique organization key
    • A display name
    • Organization admins
    • A link to a BMW IT Product or Application (if you do not have your own, use Connected AI Platform: SWP-4628)
  3. Create your repository within your organization to store and manage your project code.

Step 2: Link Your Repository via CI Trust To securely connect your GitHub repository to your Managed Kubeflow Workspace and enable CI/CD pipelines:

  1. Submit a CI Trust service request to establish a secure, automated connection between your GitHub repository and your Managed Kubeflow Workspace.
  2. Once approved, CI Trust will allow your GitHub Actions pipelines to securely authenticate and interact with your CAIP environments (clusters/namespaces) using short-lived credentials, following best security practices.
  3. For more details, see the CI Trust documentation.

This two-step process ensures your code is managed in a secure, compliant way and your automation workflows can safely access your CAIP resources.

3 - Link your CDH Environment​

To enable seamless data integration between your Managed Kubeflow Workspace and BMW's Cloud Data Hub (CDH), follow these three steps:

Step 1: Identify or Create Your Dataset and CDH Use Case/Provider

  • Determine the dataset you want to consume or provide. Datasets are managed in CDH and belong to either a CDH Use Case (for data consumers) or a CDH Provider (for data providers).
  • If you need to create a new dataset or use case/provider, follow the CDH dataset documentation and the guide to create a use case or provider.

Step 2: Create a CDH Environment

  • To access data, you must create a CDH Environment on your use case or provider. The environment acts as the technical account that will be linked to your Managed Kubeflow Workspace.
  • For details, see the CDH environment documentation.

Step 3: Link the CDH Environment to Your Managed Kubeflow Workspace

  • Once your CDH Environment is ready, you need to link it to your Managed Kubeflow Workspace. This enables your CAIP workloads to access (or provide) data in CDH securely and at scale.
  • To complete the link, follow the CDH Connection setup guide and then raise a service request to the CAIP platform team.

For more information:

This process ensures your Managed Kubeflow Workspace is securely connected to the right CDH data resources for both consuming and providing data.