Build and evaluate a fraud detection model with TensorFlow and ONNX

Learning path | 4 resources | 1 hr and 25 mins | Published on May 20, 2025

Learn how to deploy a trained model with Red Hat OpenShift AI and use its capabilities to simplify environment management. By the end of this learning path, you'll have gained familiarity with managing and deploying your models effectively using OpenShift AI. We will use fraud detection as the example use case.

Try it in our Developer Sandbox

  1. Overview: Build and evaluate a fraud detection model with TensorFlow and ONNX
  2. Create and configure your workbench Page | 10 mins
  3. Configure and execute an experiment Page | 45 mins
  4. Train and verify your model Page | 30 mins

Page

Create and configure your workbench

May 20, 2025
Alex Krikos Ramakrishna Yekulla Ian Lawson
Related topics:
Artificial intelligenceDeveloper Tools
Related products:
Developer SandboxRed Hat AIRed Hat OpenShift AIRed Hat OpenShift
TensorFlow-Onnx-LP-featured-image

To get started, we'll establish a new data science project within OpenShift AI that leverages a pre-configured TensorFlow image. This image provides a ready-to-use environment for building and training your machine learning models.

Prerequisites:

In this lesson, you will:

  • Create a workbench.
  • Configure a workbench.
  • Verify workbench status.
  • Access Jupyter Notebook.

Launch a Jupyter notebook with a TensorFlow server image

Before launching the notebook, make sure you have the Red Hat Developer Sandbox set up. If not, check here for more information.

Create a new workbench in OpenShift AI

To begin, follow these steps to create a new workbench:

  1. Launch the Developer Sandbox here, then select Red Hat OpenShift AI (Figure 1).

    The Developer Sandbox page displays the OpenShift AI launch button and other services in the navigation menu on the left and center tiles.
    Figure 1: Launch OpenShift AI in the Developer Sandbox.

  2. Log in to OpenShift AI and navigate to Data science projects in the left menu. You will find a pre-created data science project associated with your username (Figure 2).

    The Data science projects overview page shows the user’s pre-created project.
    Figure 2: Select your data science project on the OpenShift AI overview page.
  3. When you click the pre-created data science project, the overview page for the data science project will appear. On this page, the first big tile on the top left is Workbenches.

  4. In this tile, click Create a Workbench (Figure 3).

    The overview page for the pre-created data science project shows the workbenches tile.
    Figure 3: Click the create a workbench button on the data science project page.

Configure the workbench

Next we’ll configure the workbench:

  1. Provide a descriptive name for your workbench.
  2. Under Notebook image, select TensorFlow and keep the version set to the default one selected.
  3. In the Deployment size section, choose Medium for the container size.
  4. Select Create workbench to initiate the creation process and launch your new workbench environment.

Verify workbench status and access Jupyter Notebook

After selecting Create workbench, monitor the workbench status. The status should eventually transition to Running, indicating the successful creation of the workbench (Figure 4). 

To continue, follow these steps to access the Jupyter notebook:

  1. Click the Open ↗ button located next to the workbench name. 

    The Data Science Projects page shows the workbench is in running status.The open button displays to the right of the workbench name.
    Figure 4: Workbench is in Running status. The open button displays to the right of the workbench name.
  2. You will land on a log in page to authenticate yourself. Select DevSandbox. You may encounter a permission prompt. Select Allow selected permissions to receive the necessary access.

  3. After allowing permissions, you will be redirected to the JupyterLab launcher. Select Python. The Jupyter Notebook environment will come pre-configured with TensorFlow and its dependencies (Figure 5).

    The Jupyter launcher page shows TensorFlow and its dependencies.
    Figure 5: Access the Jupyter launcher page.

    You will land on a blank Jupyter notebook as shown in Figure 6.

    A blank Jupyter notebook.
    Figure 6: The Jupyter notebook is blank.

  4. Clone the following GitHub repository directly into the workbench:

    https://github.com/redhat-developer-demos/openshift-ai.git

  5. After cloning the repository, navigate to the openshift-ai/4_Models_inferencing directory and open the Fraud_detection_model_with_tensorflow.ipynb notebook. This notebook includes all the necessary code snippets, along with detailed explanations and a discussion of the results essential for completing this learning path.

    The code snippets for the rest of this learning path assume that the current directory is set to openshift-ai/4_Models_inferencing. Be sure to navigate to this directory before running the code snippets.

You’ve now successfully created and configured your workbench and accessed the Jupyter notebook. Let’s move on to using OpenShift AI to execute an experiment.

Previous resource
Overview: Build and evaluate a fraud detection model with TensorFlow and ONNX
Next resource
Configure and execute an experiment