Customized
AI Models
For your business
ellamind GmbH specializes in the development of custom-tailored, application-specific AI models with the option for “on-premise” deployment. With our solutions, you can automate the processing of your private or highly sensitive data without it ever having to leave your own network. Furthermore, our models allow complete independence and data sovereignty and can significantly outperform the performance of generic AI models for specialized applications.
Our founding team has developed some of the most advanced and widely used open-source AI models in the German-speaking world and has also attracted international attention (including several top-10 placements in the global “Open LLM Leaderboard” and the first working Mixtral implementation). We stand for rapid and advanced development of state-of-the-art AI models and pipelines and will continue to openly share many of these advancements through DiscoResearch.
Our partnership with the JAAI Group enables our team to reliably implement even large and demanding projects.
References
LeoLM
Llama2 German, LeoLM Chat, EM German, DiscoLM German
DiscoLM 70b und DiscoLM 120b
Mixtral, Mistral AI´s MoE Modell
Team
Dr. Jan Philipp Harries
Co-Founder & CEO
- Creator of “LLAMA2 Chat German” and “EM German”, the most widely used German-language AI language models
- Contributor to some of the most widely used AI software libraries (including Transformers, Langchain, Axolotl)
- PhD in economics, graduate of YC Startup School Founder Track; more than 10 years of experience in the finance and energy sectors
Björn Plüster
Co-Founder & CTO
- Initiator of LeoLM, a project in collaboration with LAION and Hessian.AI, where state-of-the-art AI models such as Llama-2 and Mistral have learned the German language through “continued pretraining” on multiple supercomputers
- Creator of one of the first German LLM benchmarks, founder of DiscoResearch, and the main person responsible for several AI models that have achieved places in the top 10 of international rankings
Advantages of custom LLMs (Large Language Models)
Privacy
On-premise deployment for use of AI models with sensitive information or client data (e.g. in Healthcare).
Resilience & Independence
Full control over business-critical AI models and processes, no dependence on any single provider.
Performance
Our AI models can outperform generalist models via Finetuning and continued Pretraining with domain- and business specific data.