The ELMTEX project delivers specialized AI solutions tailored to the European healthcare environment, with applications extending to security and e-government sectors. Our approach optimizes small-to-medium language models for domain-specific tasks, providing cost-effective alternatives to off-the-shelf commercial AI solutions.
Our Approach
ELMTEX emerged from a decade of research integrating machine learning into healthcare applications. We have developed techniques to process clinical documentation efficiently, particularly addressing the paper-based reality of many European healthcare institutions. Our three-pronged modeling approach (Naive Prompting, Retrieval-Augmented Learning, and LoRA Fine-Tuning) transforms unstructured medical text into standardized formats while maintaining privacy and security.
Our most significant finding shows that smaller, fine-tuned models often outperform larger models for specialized tasks, dramatically reducing hardware requirements and costs for healthcare providers. This makes AI implementation accessible even to institutions with limited IT budgets.