Code with Me: Run Your Polars Code on the GPU @NVIDIADeveloper

Uploaded August 2026 | Updated September 2026, 2 weeks ago

Polars is one of the fastest-growing DataFrame libraries in the Python ecosystem, and now you can run the same Polars code on the GPU with cudf-polars.

In this "Code With Me" session, Brian Tepera (PM for NVIDIA), William Hill (TME for NVIDIA), and Robin van den Brink (PM for Polars) will walk through why this matters and how to use the GPU engine for Polars.

Demo done on a real financial market dataset.

What you'll learn:

- How to switch an existing Polars LazyFrame query to the GPU engine (.collect(engine="gpu")).

- Which patterns benefit most from GPU acceleration: high-cardinality group_by, sort, join, and windowing.

- How to think about data size and cardinality when deciding between CPU and GPU Polars.

- A live look at a real FSI-flavored workflow on USA stock market data.

Join us live and ask questions throughout the session.

Code with Me: Run Your Polars Code on the GPU

$Gordon Bell Winner: Forecasting Tsunamis in Real Time With Digital Twins | NVIDIA GTC The Cascadia subduction zone is capable of unleashing a magnitude 9 earthquake with tsunamis as high as 30 meters, endangering millions in the Pacific Northwest. While it has been quiet since 1700, scientists place a 37% probability of an 8.2+ earthquake in the next 50 years. In response, the 2025 Gordon Bell Prize was awarded to UT-Austin, UC San Diego, and Lawrence Livermore National Laboratory for the world’s first digital twin for real-time, full-physics-based tsunami early warning. Leveraging exascale-class HPC systems, it delivers unprecedented accuracy and speed, achieving a ten-billion-fold performance improvement over existing state-of-the-art methods. This enables rapid, data-driven hazard assessment and actionable insights during critical windows. The work not only redefines the boundaries of real-time tsunami forecasting, but also sets a new standard for disaster resilience. Omar Ghattas | Professor and Cockrell Endowed Chair in Engineering | The University of Texas at Austin Stefan Henneking | Research Associate | The University of Texas at Austin Key Takeaways: With advanced algorithms and exascale-class computing, digital twins can achieve both full physics accuracy and real-time response. Billion-parameter inverse problems (inferring large-scale model parameters from observational data) can be solved in a fraction of a second. Industry: Academia / Higher Education Topic: Simulation / Modeling / Design - Climate / Weather / Ocean Modeling Technical Level: Technical - Beginner Intended Audience: Research: Non-Academic NVIDIA Technology: Grace CPU, Hopper #nvidiagtc$

Use the Jetson AGX Orin Developer Kit to Emulate All Six Jetson Orin Modules

Jensen Huang Surprises Peter Steinberger, Creator of OpenClaw, at #NVIDIAGTC Live

Discover How Robots Use Generative AI to Reason and Act with ReMEmbR

How to Contribute to Open Source - Live on NVIDIA NemoClaw  |  Nemotron Labs

MLOps 101: Platforms and Processes for Building AI | NVIDIA GTC

Give Codex a Deep Research Skill With NVIDIA AI-Q

How to Augment Videos at Scale With Open-Source NVIDIA Physical AI Agent Skills

Developing Local AI Copilots with LangChain, NVIDIA NIM, and FAISS | LLM App Development