GPU Mode Hackathon
September 6, 2025, New York City

Join us this fall for GPU Mode’s East Coast debut!

Jane Street is hosting a hackathon that will showcase the challenges of putting a real model dealing with market events into production. In financial markets, every optimization is essential for maintaining competitive advantage. Compete head to head to see who will win the race, but keep in mind that trades will occur throughout the day. While the fastest model is great, getting a model deployed as quickly as possible will matter just as much. The day will feature a talk by Tri Dao and a panel with some of the OG PyTorch team: Soumith Chintala, Sam Gross, and Gregory Chanan.

About the model

We care about sequence models of all kinds. Transformer models have been fairly optimized by the open source community, so the focus of this event will be on more unique architectures. Making individual kernels faster is not the only way to win – you’ll have to think about the whole system, rely on profilers to identify bottlenecks, or come up with better algorithms to deal with new market events.

Coreweave and Northflank are generously providing GPUs and a platform for easily deploying and managing your workloads on GPUs and CPUs.

Register your interest in attending by August 17th. We will notify selected attendees of the final details for the event.

While we are providing the space, participants are responsible for booking their own travel and accommodations for this event.

In partnership with

Machine Learning at Jane Street

We believe deep learning is the future of quantitative trading.