Skip to main content

Generative AI (LLMs)

Responding to Emails with LLMs

We have developed a secure, German-speaking GPT model to respond to customer inquiries more efficiently. The aim was to increase speed and quality in customer service while simultaneously reducing workload. Over the course of three months, the underlying large language model (LLM) was assessed, corrected, and continuously improved by experts. The result is an AI that can answer policyholders’ questions — from pregnancy to dental prostheses — quickly and accurately in the style of the health insurance company.

Creation and Correction of Medical Letters

We support doctors in spending less time on repetitive ancillary tasks. To this end, we are developing a specialized LLM that can correct spelling and grammar in medical letters. This particularly enables doctors whose native language is not German to work more independently. For quality assurance, we work closely with a team of doctors from a large hospital group. In later stages, we will also create continuous texts directly from notes and transcripts.

Anonymization & Pseudonymization

Data protection is our top priority. We continually enhance our methods for pseudonymization and anonymization using advanced AI technologies. These allow us to reliably identify and contextually replace even complex expressions, ensuring the original version cannot be distinguished from the modified one.

Fine-Tuning of LLMs

ellamind develops custom-tailored language models that can be installed in your own data center, offering complete independence and data sovereignty. With developers of the most successful German open-source LLMs, such as LeoLM, EM-German and DiscoResearch, ellamind has some of the best LLM developers in the world.