Automating GPU Kernels: Mako’s CEO on Crossing the CUDA Moat
Aug 27, 2025•26 min•Ep. 8
Listen to this episode in Metacast mobile app
Don't just listen to podcasts. Learn from them with transcripts, summaries, and chapters for every episode. Skim, search, and bookmark insights. Learn more
Episode description
Waleed Atallah from Mako joins us to discuss automating GPU kernel generation with AI.
03:07 – Why GPU kernels bottleneck progress
04:46 – Three current ways to get kernels
06:02 – LLMs: can they write fast kernels?
07:48 – Mako’s “kernel agent” in action
09:12 – Early adopters & best-fit users
12:05 – Data scarcity limits fine-tuning
16:18 – AI-written kernels by 2030?
20:27 – Top technical risks ahead
#BeyondCUDA #GPUkernels #mako
🌐 Connect with us:
https://www.linkedin.com/in/jtatarchuk
https://www.linkedin.com/in/waleedatallah
- (00:00) - C1750 P1315 Waleed Atallah Audio
- (26:24) - OUTRO SEQUENCE
For the best experience, listen in Metacast app for iOS or Android