Vertex AI Workbench

Access data without leaving JupyterLab

Use the Cloud Storage integration to browse files and the BigQuery integration to query tables, preview results, and load data directly into your notebook.

Streamline data exploration and analysis within your notebook environment.

Run notebooks on a schedule or on demand

Use the executor to run a notebook file as a one-time execution or on a recurring schedule. Choose the environment and hardware, and use parameters to customize each run.

Automate and scale notebook runs, even when your instance is shut down.

Share insights easily

Executed notebook runs are stored in a Cloud Storage bucket, so you can share results with others by granting access to the bucket.

Collaborate effectively by sharing reproducible notebook outputs.

Secure your instance with VPC, CMEK, and Confidential Computing

Deploy within a VPC network, use customer-managed encryption keys (CMEK) for data at rest, and enable Confidential Computing to encrypt data in use.

Meet compliance and regulatory requirements with robust security controls.

Add conda environments and custom containers

Add conda environments to use kernels like R or Apache Beam, or create an instance from a custom container based on a Google-provided base image.

Tailor your environment to specific frameworks or versions.

Integrate with Apache Spark and Compute Engine reservations

Run notebooks on a Managed Service for Apache Spark cluster, and use Compute Engine reservations to ensure VM resources are available for your instances.

Process data faster and guarantee resource availability for critical workloads.

• Jupyter-based environmentAgent Platform Workbench instances are Jupyter notebook-based development environments for the entire data science workflow. They are prepackaged with JupyterLab and include a preinstalled suite of deep learning packages, including support for TensorFlow and PyTorch frameworks.
• Integrated data accessYou can access data without leaving the JupyterLab interface. The Cloud Storage integration lets you browse data and files, while the BigQuery integration allows you to browse tables, write queries, preview results, and load data into your notebook.
• Scheduled notebook executionUse the executor to run a notebook file as a one-time execution or on a schedule. You can choose the specific environment and hardware. Notebook code runs on Agent Platform custom training, enabling distributed training, hyperparameter optimization, and continuous training jobs. Even when your instance is shut down, scheduled runs execute and save results.
• Secure by defaultAgent Platform Workbench instances are protected by Google Cloud authentication and authorization. You can deploy your instance in a VPC, and supported capabilities include Confidential Computing, customer-managed encryption keys, and service perimeter controls.

• Sharing insightsExecuted notebook runs are stored in a Cloud Storage bucket, so you can share insights by granting access to the results. The documentation does not specify any other sharing mechanisms or integrations beyond this.
• GitHub syncAgent Platform Workbench instances support the ability to sync with a GitHub repository. The documentation does not provide details on how this sync works or any limitations.
• Hardware configurationYou can configure either CPU-only or GPU-enabled instances. The documentation does not list specific machine types or GPU models available.
• PricingThe scraped content does not include any pricing information for Agent Platform Workbench. The page mentions a 'Pricing' link in the navigation, but no pricing details are provided in the scraped content.

Access Data from JupyterLab

Use the Cloud Storage integration to browse files and the BigQuery integration to query tables, preview results, and load data directly into your notebook without leaving the JupyterLab interface.

Schedule and Execute Notebook Runs

Run notebook files as one-time executions or on a recurring schedule using the executor. Choose the environment and hardware, and even run while your instance is shut down, with results saved to Cloud Storage.

Train Models with Custom Hardware

Configure CPU-only or GPU-enabled instances and run your notebook code on Agent Platform custom training to support distributed training, hyperparameter optimization, and continuous training jobs.

Secure Your Instance with Advanced Controls

Deploy within a VPC, use customer-managed encryption keys (CMEK), enable Confidential Computing for data-in-use encryption, and set automated shutdown for idle instances to manage costs.

Define the exact google Colab AI Tools workflow Vertex AI Workbench should support.

Compare it with closely related AI tools in the same category before committing.

Set review rules for accuracy, privacy, brand voice, compliance, and final approval.

Connect useful outputs to the wider stack instead of leaving them inside the AI tool.