Hessian.AI
Service Center for Artificial Intelligence
The AI service center hessian.AISC, funded by the Federal Ministry of Education and Research and located at hessian.AI in Darmstadt, strengthens the European AI ecosystem and the technological sovereignty of Europe by
- providing and expanding a massive AI computing infrastructure,
- conducting application-oriented research in the context of the 3rd wave of AI,
- offering a wide range of services, and
- providing educational training in the field of artificial intelligence.
Through these four pillars, we support the transfer of today’s cutting-edge research into the services and products of tomorrow. We aim to develop robust, secure, and efficient AI systems for a broad range of users and reduce the barriers to applying and further developing artificial intelligence.
Leveraging hessian.AI’s expertise, hessian.AISC provides a unique ecosystem to help startups, companies, and public institutions advance in the field of artificial intelligence and to drive forward cutting-edge research in Germany.
Our services at a glance
Scalable AI computing resources
Our AI Service Center provides easy and uncomplicated access to up to 632 A100 80GB SMX GPUs to train even large models quickly and carry out efficient proofs of concept. Our experts support you with their experience.
→ To find out more, take a look at our computing resources
Use and apply state-of-the-art models directly from research
hessian.AISC’s research results are available on our Model Hub. They can be used for further applications. Some of them are directly available as demonstrators, so they can be tested in an application context without much effort and high-level expertise. Our experts are happy to answer any questions regarding adaptation and application.
→ To find out more, take a look at the 3rd wave of AI
Consulting and support
Our unique ecosystem enables us to support you in developing your models. From the initial evaluation of your idea and a quick proof of concept to the complete development of an application, we are happy to support you along the way.
→ To find out more, take a look at our Services
Cooperative cutting-edge research with scientists from hessian.AI
As part of cooperative research projects, you not only gain access to our computing infrastructure, but you also have the chance to collaborate with our top researchers to solve your problems and issues.
→ To find out more, take a look at our Services
Educational training for every level of knowledge
The AI Service Center helps companies and institutions become future-ready by offering a wide range of workshops, seminars, lectures, and online courses to build essential theoretical and practical AI knowledge.
→ To find out more, take a look at our training programs
Events
GRAG – German Retrieval Augmented Generation
Large language models can only utilize their full potential when they have reliably access to up-to-date, organization-specific knowledge bases. This is exactly where GRAG comes in: In close collaboration with Avemio AG, hessian.AISC has developed an open-source model suite that, consistently tailors Retrieval Augmented Generation to the German-speaking society.
The suite includes specialized language models ranging from 4 to 12 billion parameters, embedding models for semantic search, and a speech-to-text model for processing spoken content—all trained on over 3.1 million curated training examples using the hessian.AI supercomputing infrastructure. All models can be run entirely on-premises, thereby giving companies, government agencies, and research institutions full control over their data.
GRAG serves as an example of how hessian.AISC facilitates the transfer of cutting-edge AI research into practical applications: openly, confidently, and tailored to the needs of the German-speaking society. Models, datasets, benchmarks, and a self-hosted starter kit are freely available to the community.
Funded by the Federal Ministry of Research, Technology and Space (BMFTR).
Models on Hugging Face · Starter Kit on GitHub · GRAG Training Study (PDF)
