Skip to content
MeetupOnlinePast event

GPU, CUDA, and PyTorch Performance Optimizations

Explore the topics

Event artwork for GPU, CUDA, and PyTorch Performance Optimizations

This event has ended

This listing is kept as an archive. Explore upcoming events to find your next one.

See upcoming events
About this event

What to expect

Master GPU acceleration and performance optimization for deep learning in this comprehensive online workshop focused on CUDA and PyTorch. Whether you're struggling with slow model training times or looking to squeeze every ounce of performance from your hardware, this session delivers practical techniques for optimizing AI workloads at the GPU level.

What You'll Learn

This meetup dives deep into the performance engineering aspects of modern AI systems, covering GPU architecture, CUDA programming fundamentals, and PyTorch optimization strategies. The session features multiple technical talks exploring how to identify bottlenecks in your training pipelines, implement efficient memory management patterns, and leverage CUDA kernels for maximum throughput. You'll discover practical approaches to profiling GPU utilization, optimizing data loading, and implementing mixed-precision training.

The event draws from real-world examples documented in the accompanying GitHub repository and the O'Reilly book "Systems Performance Engineering: Optimizing Algorithms" — giving you access to production-tested code samples and reference implementations you can apply immediately.

Who Should Attend

This session targets ML engineers, data scientists, and researchers working with deep learning frameworks who need to improve model training performance. While prior experience with PyTorch is helpful, the content is structured to accommodate both intermediate practitioners looking to level up their optimization skills and experienced engineers seeking to refine their performance tuning methodology. If you're training large models, working with limited GPU resources, or simply curious about what's happening beneath the high-level API calls, you'll find actionable insights here.

Format and Resources

The meetup runs via Zoom with live technical presentations followed by Q&A. The recurring monthly series (every third Monday) creates an ongoing learning community focused on AI performance engineering. Attendees gain access to supplementary materials including the YouTube channel with recorded sessions and a free course on Generative AI through DeepLearning.ai for those looking to expand their foundation.

Hosted by Chris Fregly and Antje Barth, who bring extensive experience in production ML systems, the event combines

Running an AI event?

List it here, free.

Meetups, workshops and conferences go in front of thousands of builders after a quick review.

List your event