Our mission is to build a platform that allows data scientist to explore data, train machine learning algorithms, and build applications while primarily staying on the GPU


The GPU Open Analytics Initiative (GoAi) seeks to foster and develop open collaboration between GPU analytics projects and products to enable data scientists to efficiently combine the best tools for their workflows.

Why This is Necessary

Most GPU-enabled software treats the GPU as an implementation detail to be hidden from their external interfaces. This is a sensible choice when only one project in a data science pipeline is GPU-accelerated, but it becomes less and less efficient as more applications and libraries in the pipeline gain GPU-accelerated implementations. The GoAi members recognized this difficulty, and are collaborating to create an open spec and set of tools for data exchange between libraries and applications without needing to move data off the GPU.

We want to make accelerated, end-to-end GPU analytics easy.


There are many ways for a project, group, individual, or company to join the GPU Open Analytics Initiative (GoAi). Below are details. If you have any questions, please reach out.

Slack Workspace
Google Group

Learn, try, and collaborate with us.

Become an Adopter

Adopters are projects, products or groups that are currently using or planning to use a GoAi project. For example, a database adding support for the GPU Data Frame can be a GoAi adopter.


Adopters should announce their plans on our public Google Groups, and we will add a link to their project on the GoAi website.

Become a Contributor

Contributors are any projects, groups, or individuals that would like to help develop projects. Since the GoAi is based around open projects, clone a repository on our GitHub channel and make a pull request. To help faciliate the process, let us know of your plans on our public Google Groups.


Our contributors are:

Apache Arrow, Simantex, Siu Kwan Lam, Arno Candel, Minggang Yu, Stanley Seibert, Jon Mckinney, Bill Maimone, Vinod Iyengar, Todd Mostak,

Become a Member

A member is responsible for stewardship of the GoAi organization, including decision-making about expanding GoAi to future projects, planning for events and outreach, and generally promoting the GoAi mission. Members will meet periodically to discuss and vote on GoAi management issues. However, technical decisions are handled by the open source developer communities around individual GoAi projects and membership is not a requirement for technical contributions. An organization can request membership from the existing members.

If the criteria is applicable, apply for membership on our public Google Groups. Existing members will vote on the admission of new members based on the above criteria. Note that adopters do not need to satisfy the above requirements.


Members are expected to follow the below criteria:

  1. Have demonstrated contributions toward open GPU analytics.
    This can be in the form of open source code contributions, community involvement, education, etc. (Members do not have to be exclusively open source, of course.)
  2. Engage constructively with the community.
    Members should have a constructive and friendly attitude toward others in the GoAi community and be willing to mentor newer contributors and users.
  3. Collaborate with other GoAi members.
    Although GoAi members may be business competitors, within the context of GoAi meetings and projects, members are expected to collaborate in good faith with other members.

Below are our current members:


Our projects are based around an open platform, which is the foundation of the GPU analytics ecosystem. Members will help guide the development of projects, but anyone can be a contributor. And because its open, anyone can freely use the projects.

Currently, GoAi is supporting the GPU Data Frame project (GDF), which will enable tabular data to be directly exchanged between libraries and applications on the GPU. We expect other projects to be added in the future, such as a graph standard for the GDF.

We are developing our platform to be open, accessible, and available for broad adoption.

GPU Data Frame (GDF)

The basic approach for the GPU Data Frame (GDF) is pretty simple: if applications and libraries agree on an in-memory data format for tabular data and associate metadata, then just a device pointer to the data structure need be exchanged. Additionally, the IPC mechanism built into the CUDA driver allows device pointers to be moved between processes.

The GDF uses the Apache Arrow columnar format to represent data on the GPU. Some Arrow features are not yet supported.

Try our iPython Notebook Demos
Learn more at the Github Technical Overview
Ask questions on our Slack team
Or resort to StackOverflow

GDF Graph

In the planning stages.