Skip to content
VirtualOnline

Make Your GPU Earn Its Keep: A Kernel Engineering Deep Dive

Explore the topics

Event artwork for Make Your GPU Earn Its Keep: A Kernel Engineering Deep Dive
About this event

What to expect

Explore how to diagnose slow GPU workloads and engineer faster AI kernels in this online technical session on 30 November 2026, from 17:00–18:00 UTC.

Dr Rajat “Raj” Dandekar, co-founder of Vizuara and an instructor of its Kernel Engineering course, will present a practical framework for identifying GPU bottlenecks, choosing what to measure and verifying whether an optimisation has worked.

What the session covers

  • Connecting the roofline model with CUDA execution and GPU memory hierarchy
  • Profiling AI workloads to locate meaningful bottlenecks
  • Coalesced memory access, tiling and tensor-core use
  • Matrix multiplication and attention as real-world optimisation contexts
  • Modern kernel techniques, including FlashAttention

The programme begins with meetup updates from hosts Chris Fregly and Antje Barth, followed by Raj’s dedicated technical talk and time for audience questions. It is aimed at engineers comfortable with Python or basic PyTorch; previous CUDA or C++ experience is not required.

This event is hosted online by the AI Performance Engineering Meetup (Amsterdam). The listed time is 09:00–10:00 PST / 17:00–18:00 UTC / 18:00–19:00 in Amsterdam. The Zoom link is available on the event page for attendees.

Scheduling note: 30 November is a proposed new date, and speaker confirmation is still pending.

Register for the event on Meetup

Running an AI event?

List it here, free.

Meetups, workshops and conferences go in front of thousands of builders after a quick review. Paste the link, add your images, and we will email you when it is live.

List your event