Learn about the project, use cases,
and engagement with RAPIDS

About

RAPIDS is a collection of open source software libraries and APIs that gives you the ability to execute end-to-end data science and analytics pipelines entirely on NVIDIA GPUs using familiar PyData APIs. Jump To Section

Use Cases

Learn more about RAPIDS success stories, use cases, applications, research and other topics related to GPU accelerated data science. Jump To Section

Get Involved

Get in contact with RAPIDS, find enhanced support with NVIDIA enterprise services, and reach out to the broader developer community. Jump To Section

About The Project

Familiar

Utilizing NVIDIA CUDA primitives for low-level compute optimization, RAPIDS exposes GPU parallelism and high-bandwidth memory speed through user-friendly interfaces:


Dataframe processing with cuDF (similar API to pandas)
Machine learning with cuML (similar API to scikit-learn)
Graph processing with cuGraph (similar API to networkX)
Spatial analytics with cuSpatial (similar API to geoPandas)
Image processing with cuCIM (similar API to scikit-image)
Seamless cross-filtered dashboards with cuxfilter
Low level compute primitives with RAFT
Apache Spark acceleration with Spark RAPIDS

and many more projects...

Scalable

RAPIDS delivers impressive acceleration from on a single GPU desktop system all the way to MNMG (multi-node multi-GPU) configurations on an expansive range of NVIDIA hardware.


Impressive single GPU performance out of the box
MNMG with RAPIDS + Dask and RAPIDS + Spark
HPC with RAPIDS + Slurm

Open Source

RAPIDS had its start from the Apache Arrow and GoAi projects based on a columnar, in-memory data structure that delivers efficient and fast data interchange with flexibility to support complex data models.


With NVIDIA backing, it has maintained a focus on open source development (typically Apache 2.0 licensed) and collaboration with a broad community of cutting edge data science projects.

PyData Accelerators

cuDF Pandas Accelerator

cuDF has a new pandas accelerator mode to speed up pandas workflows with zero code change. This is part of an on-going effort to streamline accelerated data science. Learn More on the Launch Page

Polars GPU Engine

RAPIDS and Polars are collaborating to GPU accelerate Polars DataFrames, enabling workflows to utilize both the GPU and CPU to significantly speed up workloads. Learn More on the Launch Page

NetworkX Accelerator

The cuGraph backend for NetworkX brings accelerated graph analytics to every NetworkX user. This is another part of an on-going effort to streamline accelerated data science. Learn More on the cuGraph Docs Page

Enterprise Use Cases

Walmart + RAPIDS ML

When you are a global-scale company, precision is everything. Learn how Walmart implemented XGBoost to increase forecast accuracy and potentially save billions of dollars. Read More

Bumble + RAPIDS ML

Buzzwords is Bumble's open-source GPU-powered topic modelling tool, developed in house and building upon the work found in BERTopic and Top2Vec. Buzzwords uses cuML's UMAP and HDBSCAN algorithms. Read More

AT&T + RAPIDS ETL

AT&T was able to do more, faster by moving data science workloads to GPU. Hear about their analysis and use of NVIDIA capabilities across different domains and share specific use case examples, their efficiency gains, and corresponding impacts. Hear More

Amazon + RAPIDS GNN

See how RAPIDS is enabling work on the very cutting edge by enabling Graph Neural Networks (GNN) for applications including drug discovery, recommender systems, fraud detection, and cybersecurity. Hear More

More Industry Use Cases...

Find more resources for industry specific ML, AI, and data science uses cases on NVIDIA's industry page. Read More

Data Science Research

KGMON + RAPIDS

Learn about Kaggle Grandmasters of NVIDIA (KGMoN), and see how they use RAPIDS to build winning recommender systems, predict degradation rates in RNA molecules, identify melanoma in medical imaging, and more. Read More on the KGMON Page

Research Papers

Using RAPIDS in your data science research and development? Let us know for opportunities to promote it. See the RAPIDS Citation Guide

Get Involved

Business

Use RAPIDS directly or through NVIDIA AI Enterprise, which provides extensive optimization, certified hardware profiles, and direct IT support. Read more about NVIDIA AI Enterprise

Community

Reach out and engage with the RAPIDS community on the following channels:

Slack Channel
Stack Overflow
X/Twitter
NVIDIA Developer Blogs

Evangelism

Promote and foster more use of RAPIDS with these Deep Learning Institute (DLI) and Launchpad courses:

Accelerating End-to-End Data Science Workflows
Fundamentals of Accelerated Data Science
NVIDIA LaunchPad Data Science Lab