yasp.ai is an AI infrastructure company that makes machine learning models run fast on any hardware. Its AI-driven compiler generates and caches optimized compute kernels automatically, delivering 2–10× faster training and inference through a single API call — across NVIDIA, AMD, Intel, Qualcomm and Apple silicon, from datacenter GPUs to edge devices. Founded 2023 in Munich.
yasp benchmarked three models with distinct computational profiles — RegNet, DeepSeekV3 MLA and a MiniGPT Block — on Azure’s AMD Radeon PRO V710 GPU against torch.compile on an Nvidia A10. yasp.compil
…
AI-written GPU kernels have been winning on isolated operations for over a year, but those wins rarely survive the trip to a real model — Google’s AlphaEvolve moved Gemini training by roughly 1%. Benc
…
Benchmarking AI inference on Azure, yasp’s agentic compiler delivered a 2.91× speedup for the MiniGPT Block on AMD’s Radeon PRO V710 GPU over torch.compile. This technical deep-dive breaks down exactl
…
3 Results
Loading
Contact Exhibitor
Share
Website Search
Wishlist
Add some favourites to your wishlist to get started!