264001
eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJid2V2ZW50cy1wdWJsaWMtYXBpIiwiYXVkIjpbInB1YmxpYy13ZWIiXSwic3ViIjoiaHR0cHM6Ly93d3cuZ29vZ2xlY2xvdWRldmVudHMuY29tIiwiaWF0IjoxNzc3Mzg0MDExLCJuYmYiOjE3NzczODQwMTEsImV4cCI6MTc3NzM4NzYxMSwiY29udGV4dCI6eyJwYXlsb2FkIjp7ImlkIjoiIiwicmVnVHlwZU5hbWUiOiIiLCJldmVudF9pZCI6IjI2NDAwMSIsImNfMzI5MDQxNiI6IiJ9fX0.fRTmQbBhtELj655uf5WEAGbFow7VE07rKee6F73mqBA
phase 7 playlists session info modal
Day 2 – April 23, 2026
Discussion Groups
BRK1-068B-DG • Compute • Advanced Technical
Max performance on TPUs: Kernels, XLA, and SparseCore
location_on Reef A
schedule 2:45 PM - 3:15 PM

How do you extract maximum ROI from your hardware? We'll discuss the infrastructure driving the next generation of AI speed. This discussion will explore the latest in XLA compilers, SparseCore optimizations, and custom kernels. Share your strategies for reducing TCO and accelerating model execution. Whether you are hacking on Mosaic or tuning with Tokamax, join fellow performance engineers to swap secrets on pushing the AI Hypercomputer to its absolute limit.

 Read more
This agenda widget contains logic for handling same sessions available at multiple times
GTM Tracking Header name present
Session
contains logic for gtm tracking when modal open