GPU, CUDA, and PyTorch Performance Optimizations

Description

This monthly online meetup brings together AI developers and engineers focused on squeezing maximum performance from GPU-accelerated deep learning workloads. Hosted by Chris Fregly and Antje Barth, the session dives deep into the practical optimization techniques needed to make PyTorch models run faster and more efficiently on GPU hardware. The September 2026 meetup features two technical talks specifically focused on GPU, PyTorch, and CUDA performance optimizations. Whether you're training large language models, running inference at scale, or simply trying to reduce your cloud GPU bills, understanding these low-level optimization techniques is essential for production AI systems. The meetup covers the full stack — from CUDA kernel optimization to PyTorch-specific performance patterns that can dramatically improve throughput and reduce latency. This is part of a recurring monthly series that meets every third Monday, creating a consistent learning community for performance-focused AI engineers in [Madrid](https://www.createwith.com/cities/madrid) and beyond. The online format means you can join from anywhere with an internet connection, making it accessible to a global audience of performance engineers. Chris Fregly is the author of an O'Reilly book on [AI Systems Performance Engineering](https://www.amazon.com/Systems-Performance-Engineering-Optimizing-Algorithms/dp/B0F47689K8/), bringing deep expertise in optimizing AI algorithms and infrastructure. The meetup connects to broader resources including a [GitHub repository](http://github.com/cfregly/ai-performance-engineering/) with code examples, a [YouTube channel](https://www.youtube.com/@AIPerformanceEngineering) with recorded sessions, and a free [Generative AI course on DeepLearning.ai](https://bit.ly/gllm). Ideal for machine learning engineers, DevOps professionals working with AI infrastructure, and anyone running PyTorch workloads who wants to understand performance bottlenecks at the GPU level. Prior experience with PyTorch and basic understanding of GPU computing concepts will help you get the most from the technical discussions. The meetup format includes introductions, community updates, and two focused technical talks with time for Q&A. Join via Zoom to learn from practitioners who are solving real-world performance challenges in production AI systems. [Register for this event](https://www.meetup.com/ai-performance-engineering-meetup-madrid/events/ckvnttyjcmbcc/?eventOrigin=group_upcomin