Expertise is not in the document.
It is in the model itself.
General-purpose language models often answer health questions by consulting documents loaded into them. In these models, expertise sits not inside the model but in the uploaded text.
MINA's strength, by contrast, comes from the model itself. We built Turkish clinical language into MINA's weights through continued pre-training, and added expertise in 14 specialties to the model with LoRA adapters. MINA applies regulations through rules structured with their effective dates, and these rules are monitored and kept up to date every day.
The same model answers the physician's question in the clinical assistant, prepares provision, billing and audit decisions on the TIS and SIT platforms, and reads each patient's digital twin in the data warehouse. A single model works behind all of these tasks.
On a set of 1,451 regulation and rule questions, MINA answered 98.1% of the questions correctly. It based every answer on a source and responded in 4.6 seconds on average.
You can read in detail how MINA was built and why we chose this approach on the MINA page. If you would like to see it in your own institution, you can request a POC.
MINA results come from an internal evaluation; synthetic scenarios, October 2026.
Clinical intelligence that sees the unseen and anticipates what comes next.