Skip to main content Skip to main navigation

Soofi Trainer

Application-specific, sovereign AI models for your own applications

Soofi Trainer is an AI-powered assistant that helps companies customize sovereign language models to their industry and specific use cases – transparently, responsibly, and entirely under their own control.

Small and medium-sized businesses often face challenges when adopting AI technology: Commercial providers offer powerful models, but no one knows exactly what data they were trained on or what background data is processed. Soofi Trainer addresses this issue by offering an alternative: Companies can train a model directly using their own domain-specific data for clearly defined tasks – if such data is available internally. This keeps them in control of what the model learns and how it functions.

Even companies without their own datasets benefit: Through modern data rooms – digital marketplaces where companies can securely share and purchase data under fair terms –industry-relevant data can be integrated directly into the training process. Soofi Trainer supports this approach, making the emerging data economy accessible to companies of all sizes.

The demonstrator acts like a digital advisor: Users specify their industry and use case in a dialogue – using voice input and output if preferred – and the Soofi Trainer guides them step by step through the entire process: from analysis and method selection to training execution. No prior technical expertise is needed.

The result is a tailor-made AI model that can be deployed in-house, with its development fully documented and traceable – essential for the trustworthy use of AI in sensitive fields like manufacturing, medicine, or law. 

The Soofi (Sovereign Open Source Foundation Models) consortium project is led by the German AI Association and funded by the Federal Ministry for Economic Affairs and Energy.

Visit us

Hall 13, Stand C24

More information

Contact

Dr. Daniel Porta  
DFKI Research Department Cognitive Assistants