NVIDIA and Polars Host Live Coding Session to Run Polars on GPUs with cuDF

NVIDIA and Polars held a live coding session on July 31 to demonstrate how to accelerate existing Polars workflows with the cuDF Polars GPU engine. The session, announced via NVIDIA AI's official X account, featured Brian Tepera and William Hill from NVIDIA, along with Robin van den Brink from Polars, to demonstrate a workflow using real US stock market data. The announcement was made through a post on X titled "Code with Me: Run Your Polars Code on the GPU," which directed audiences to a live stream discussing the integration between Polars, a popular Python based data analysis library, and cuDF Polars, a GPU engine developed by NVIDIA to accelerate query execution. During the session, participants were taught how to enable GPU execution with the .collect(engine="gpu") command.

This key step allows Polars users to leverage the parallel processing power of GPUs without rewriting their entire code. The session also covered identifying workloads that could benefit from GPU acceleration, particularly group by , join , sort , and window operations. These four types of operations are known to be computationally intensive components in data analysis, and accelerating them can have a significant impact on overall processing time. Participants also received a comparative evaluation between CPU and GPU execution. This comparison is important for understanding when GPU usage provides real advantages and when CPU execution remains the more efficient choice, depending on data characteristics and query complexity.

The collaboration between NVIDIA and Polars reflects a broader trend in the data computing ecosystem, where technology companies are increasingly seeking to bridge the gap between the ease of use of Python based data analysis libraries and the need for higher processing speeds. Polars has been known as a faster alternative to pandas for many data frame operations, largely due to its architecture designed to efficiently utilize multi threading. With cuDF Polars, NVIDIA takes this capability to the next level by leveraging the thousands of cores available on GPUs. This approach differs from previous efforts that required users to learn new APIs or significantly rewrite their code. By changing just one parameter in the .collect() call, existing Polars users can try GPU execution without altering their code structure.

Such live coding sessions are also part of NVIDIA's strategy to expand adoption of their computing platform beyond traditional domains like AI model training. By targeting users of mainstream data analysis libraries, NVIDIA aims to show that GPU acceleration is relevant not only for heavy AI workloads but also for everyday data analysis. The use of US stock market data as an example in this session provides relevant context for professionals in the financial and fintech industries. Stock market data typically has large volume and requires real time analysis, making it an ideal candidate to demonstrate the benefits of GPU acceleration in real world scenarios.

For the financial industry and companies that rely on high speed data analysis, the ability to accelerate group by , join , sort , and window operations on large scale data can reduce the time needed to derive insights from data. This has the potential to impact time sensitive decision making, such as risk analysis or identifying trading opportunities. From an ecosystem perspective, this collaboration also shows the increasingly close integration between NVIDIA as a provider of computing infrastructure and open source projects widely used by the data science community. Polars, developed under an open source license, is an example of how community projects can be integrated with a larger commercial ecosystem. NVIDIA's investment in developing cuDF Polars demonstrates the company's long term commitment to the data computing market.

This aligns with NVIDIA's efforts to expand beyond artificial intelligence into high performance computing and general data analytics. For developers who want to try this technology, the live coding session provided practical guidance directly from engineers involved in the development of both projects. The presence of Robin van den Brink from Polars in the session also signals official support from the Polars development team for the integration with cuDF. Looking ahead, the success of such integrations could pave the way for broader GPU adoption in mainstream data analysis tooling. If Polars users experience significant benefits from GPU execution, this could encourage other libraries to follow suit, accelerating the industry's transition toward heterogeneous computing that utilizes both CPUs and GPUs.

This live coding session is part of NVIDIA's Code with Me program series, designed to bring their technology closer to developers. Through a live demonstration format, NVIDIA aims to reduce adoption barriers by providing concrete examples that participants can immediately put into practice.