yasp, a DeepTech startup out of Munich developing the Agentic AI Compiler, announced the successful closing of its €4.2 million Seed funding round to bring development one step closer to a Q4 2025 market launch.
The funding round was led by Capnamic. Joining them was Start-up BW Innovation Fonds & MBG BW, alongside several experienced business angels.
“We are building a product to directly address a critical and growing bottleneck in AI development: the time-consuming process of model training and deployment. Our vision is to empower developers and enterprises to achieve faster results, without sacrificing flexibility or accuracy,” says Stefan Krassin, Co-founder and CEO of yasp.
According to data provided by yasp, global investment in AI is projected to reach $1.8 trillion by 2030, presenting enterprises with increasing challenges in scaling AI development. Today, more than 60% of corporate AI budgets are allocated to software and infrastructure. As models grow more sophisticated, future development is expected to require even greater financial investment.
In addition, developers face the persistent issue of vendor lock-in, limiting long-term innovation strategies. Consequently, efficiency and flexibility have become critical challenges for organisations striving to stay competitive, driving urgent demand for new solutions that accelerate AI innovation while controlling costs.
Founded in 2023, yasp reportedly addresses these challenges with their Agentic AI Compiler: a technology designed to accelerate AI model training and inference, improve performance, and reduce time-to-market. Seamlessly integrating with widely used frameworks like PyTorch, yasp’s technology transforms the way developers build and deploy AI models by making advanced hardware optimisation as simple as adding a few lines of code.
With a single API call, engineers unlock hardware-specific compilation and optimisation of AI models – with a significant difference to traditional runtime compilation: yasp generates custom kernels tailored to the target hardware, and scales seamlessly from individual kernels to entire models.
The closure of this Seed round for yasp showcases a growing appetite among European investors for startups tackling DeepTech challenges at the intersection of hardware and AI.
While many related flows in 2025 have centered on adjacent domains – chip cooling (Corintis, Switzerland, €20 million), photonic AI chips (Arago, Paris, €22.1 million), or autonomy stacks in rail (Futurail, Germany, €7.5 million) – yasp’s raise underscores that infrastructure layers closer to model compilation and runtime optimisation are garnering attention as well.
From a German perspective, the overlap with Futurail suggests that capital is active in the domestic DeepTech scene, which may aid yasp in network proximity and investor familiarity.
“Our approach is fundamentally different: instead of requiring developers to rewrite their complex models. With one API call, yasp rewrites and optimises a model’s training and inference layers for the user’s target hardware,” explains Reza Rahimi, CTO of yasp.
With recently opened offices in Munich and Montreal, yasp will use the fund to accelerate product development, expand support for additional hardware platforms, and deepen integrations with popular AI development tools.
Planning to launch the Agentic AI Compiler at the end of the year, the company is committed to helping developers and enterprises unlock new possibilities in AI innovation while controlling costs.
“yasp is redefining what is possible in AI development by bridging the gap between software innovation and hardware performance. Their Agentic AI Compiler empowers engineers to achieve unprecedented speed and efficiency, while maintaining flexibility and avoiding vendor lock-in. We are excited to support yasp as they unlock new levels of productivity and scalability for the entire AI ecosystem,” says Christian Siegele, Managing Partner at Capnamic.
Read the orginal article: https://www.eu-startups.com/2025/09/german-deeptech-startup-yasp-raises-e4-2-million-to-accelerate-ai-model-optimisation/
