yasp.ai
Category: AI Infrastructure
yasp is developing an AI compiler to accelerate the training and deployment of AI models on any hardware. yasp.ai was founded in 2023. The company is led by Stefan Krassin. Based in Munich, Germany. Team size: 11-50. Total funding raised: €4.2 million ($5 million). Latest round: Seed. Key investors include Capnamic, Start-up BW Innovation Fonds, MBG BW.
- Founded
- 2023
- Headquarters
- Munich, Germany
- Team size
- 11-50
- Total funding
- €4.2 million ($5 million)
Value proposition
The Agentic AI Compiler accelerates model training and deployment by automatically generating hardware-optimized executables with a single API call, reducing compute costs and avoiding model rewrites.
Products and solutions
Agentic AI Compiler (yasp.compile), Gaia — autonomous multi-vendor kernel/inference optimization, Kronos — deterministic PyTorch-to-native binary deployment (NVIDIA edge/cloud), yasp.agent — conversational/programmatic assistant for compilation
Unique value
The Agentic AI Compiler performs hardware-specific code generation to create optimized executables for target hardware via a single API call, differentiating it from conventional runtime compilation.
Target customer
AI developers, data scientists, and companies building and deploying AI models.
Industries served
Any industry leveraging AI and machine learning models
Technology advantage
Automated, hardware-agnostic model optimization reduces manual tuning and specialized hardware knowledge, lowering costs and accelerating AI development and deployment.
How they differentiate
yasp differentiates with its Agentic AI Compiler that simplifies hardware optimization to a single API call, is hardware-agnostic to avoid vendor lock-in, and automatically optimizes models for specific hardware without manual tuning.
Main competitors
Modular, OctoML, Deci AI
Key partnerships
AMD — case study / benchmarks on Radeon PRO V710 (e.g. MiniGPT cost reduction vs NVIDIA A10), IBM — Granite 4.0 benchmarks (company now cites up to 9x inference vs torch.compile on H200), Microsoft Azure — AMD Radeon PRO V710 / MiniGPT collaboration, NVIDIA Inception program, Agentic AI Foundation (AAIF) under Linux Foundation — member since 2026, EDGE AI Foundation — Strategic Partner, PyTorch Foundation — Silver Member, Linux Foundation — member, HAI Foundation — member, Amazon / AWS Trainium — validated hardware target
Notable customers
Torc Robotics — named impact study: autonomous trucking / perception deployment (2-4x speedup, ~30% cost reduction)
Major milestones
Closed €4.2M / $5M seed led by Capnamic (Sep 25, 2025)., NVIDIA Inception member; EDGE AI Foundation Strategic Partner; PyTorch Foundation Silver Member., yasp.agent early access (Dec 2025); product surfaces Gaia + Kronos publicly positioned (2026)., IBM Granite 4.0: company cites up to 9x inference vs torch.compile on NVIDIA H200 (Mar 2026+)., AMD Radeon PRO V710 / Azure benchmarks; MiniGPT cost reduction vs NVIDIA A10 (Apr 2026)., Named Europe's Top 100 Cloud Challengers 2026 (Notion Capital / Google Cloud)., Joined Agentic AI Foundation / Linux Foundation ecosystem (2026)., New York presence added alongside Munich and Montréal; ~25 engineers across three cities., Series A fundraising in progress (stated on about page as of Aug 2026)., Public impact study with Torc Robotics (autonomous driving).
Growth metrics
Not publicly available.
Market positioning
Positioned as a developer-first tool for enterprises in the AI infrastructure market, aiming to reduce complexity and cost of scaling AI by providing a seamless integration layer.
Geographic focus
Europe and North America — teams/offices in Munich (DE), Montréal (CA), and New York (US).
Patents and IP
No publicly available information on patents.
About Stefan Krassin
Stefan Krassin was previously the Managing Director at PLC2 Design GmbH, progressing from Sales Manager to Head of Sales and Marketing and then Managing Director, giving him experience in deep tech leadership and business development.
More AI Infrastructure companies
Official website: https://yasp.ai/