eScience 2026 Tutorial

Accelerated Data Science with RAPIDS

This site hosts the web materials for the eScience 2026 tutorial on GPU-accelerated data science. It combines the current tutorial proposal, the hands-on notebook links, and the practical information attendees need before and during the session.

Event details

• Event: eScience 2026
• Tutorial: Accelerated Data Science with RAPIDS
• Presenter: Didier Barradas-Bautista
• Affiliation: KAUST Visualization Core Lab
• Date: TBD
• Location: TBD

What this tutorial covers

Participants will work through GPU-accelerated data science workflows with the NVIDIA RAPIDS ecosystem, including:

• Data processing and feature engineering with cuDF
• Machine learning workflows with cuML and GPU-enabled XGBoost
• Graph analytics with cuGraph
• A compact advanced extension on distributed hyperparameter optimization with Ray Tune and selected Skorch/PyTorch patterns
• Practical guidance for reproducible, performance-aware data science in cloud-ready notebooks

Site navigation

• Setup instructions
• Syllabus and schedule
• Tutorial 1: Scikit-learn + RAPIDS
• Tutorial 2: Ray Tune + RAPIDS
• Tutorial 3: Skorch + RAPIDS
• Resources
• FAQ

Tutorial materials

The hands-on materials are based on the Google Colab notebooks used in the tutorial:

1. Tutorial 1: Hyperparameter optimization with Scikit-learn and RAPIDS
2. Tutorial 2: Hyperparameter optimization with Ray Tune and RAPIDS
3. Tutorial 3: Deep learning workflows with Skorch and RAPIDS

The source repository for the materials is available at D-Barradas/RAPIDS_HPO.