> This page is for Kumo Engine.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# Available Deployment Options

> Choose a notebook or a local Linux, macOS, or Windows (WSL2) environment.

Run Kumo from a notebook in Snowflake or Databricks, or from Python on Linux, macOS on Apple silicon, or Windows through WSL2.
Choose the environment that fits how you explore data, develop models, and run your workflows.
See [GPU Sizing](/sdgm/kumo-engine/get-started-with-kumo-engine/gpu-sizing).

## Notebook Deployment

Use a notebook when you want to develop your Kumo workflow interactively in the platform where you already work with data.
You can organize the workflow into cells, inspect results as you go, and keep explanations alongside your code.
This makes notebooks useful for exploring a prediction problem and sharing the steps behind an experiment with your team.

### Snowflake Notebook

Choose a Snowflake notebook if your source tables are in Snowflake and your team already uses Snowflake for data analysis.
You can develop your Kumo workflow in a familiar notebook environment, with your analysis and modeling code together.
This is a natural starting point for moving from exploring Snowflake tables to building predictive models from them.

Start with [Snowflake Notebook Setup](/sdgm/kumo-engine/get-started-with-kumo-engine/installation/snowflake-notebook).

### Databricks Notebook

Choose a Databricks notebook to train on attached GPU compute using data from Amazon S3, Azure Data Lake Storage, or Unity Catalog. Use local storage for graph and model work and a Unity Catalog volume for retained model files.

Start with [Databricks Notebook Setup](/sdgm/kumo-engine/get-started-with-kumo-engine/installation/databricks-notebook).

## Local Machine Deployment

Choose a local machine when you want to run Kumo from Python in your own environment.
macOS supports Apple GPU execution through MPS and CPU execution.
Windows uses the Linux runtime inside WSL2, with NVIDIA CUDA or CPU execution.
For production workloads, use a compatible NVIDIA GPU.
You control the Python environment, compute resources, and working storage, and can organize your workflow as scripts in your own project.

This option fits teams that want to develop with their usual editor and version control tools, work with local data files, or incorporate Kumo into an existing Python workflow.
You are responsible for preparing the machine, installing the required dependencies, and providing access to your data.

Choose your setup guide:

* [Linux Machine Deployment](/sdgm/kumo-engine/get-started-with-kumo-engine/installation/linux-machine-deployment)
* [macOS](/sdgm/kumo-engine/get-started-with-kumo-engine/installation/macos)
* [Windows (WSL2)](/sdgm/kumo-engine/get-started-with-kumo-engine/installation/windows-wsl2)