AI progress should not be bottlenecked by hardware. yasp is a model and inference optimization company. Founded in 2023, we are a German-Canadian startup with teams in Munich, Montreal and Palo Alto. Bring a trained PyTorch model and a hardware target, from NVIDIA and AMD GPUs to AWS Trainium, in the cloud or at the edge. yasp’s AI agents return the fastest version of that model, ready to deploy and checked against the original. No hand-written kernels, no dedicated inference team, no vendor lock-in. Why it matters: most AI models run far below what their hardware can deliver. Optimized for the chip, the same GPU serves more users: up to 2.3× more tokens per second than standard vLLM in our Model Catalog. That means a lower cost per token, fewer new GPUs, new models in production in days instead of months, and the freedom to choose hardware on price, not toolchain. We work with AI-native companies, inference providers, chip makers and teams building with open-weight models. See our open benchmarks: yasp.ai/model-catalog
| Website | https://yasp.ai/request-access |
| Employees | 30 (25 on RocketReach) |
| Founded | 2023 |
| Industry | Software Development |
Looking for a particular yasp employee's phone or email?