The Spanish generative and open source artificial intelligence (AI) model, Alia, will have its first public application before the end of the year. The tool, which will serve as an AI assistant for doctors focused on heart diseases, will be trained with medical records from thousands of patients to offer personalised responses. The heart monitor app, named Cardiomentor, is being developed by the research center Tecnalia and the Barcelona Supercomputing Center (BSC) in collaboration with the Spanish Society of Cardiology.

In the first phase, currently underway, Cardiomentor will rely on Alia’s foundational models to develop a digital helper specialised in heart diseases for general practitioners, with the goal of providing them with easy access to updated and accurate knowledge about these conditions. Once this phase is completed, the system will be trained with anonymized patient data to refine the tool’s predictive and diagnostic capabilities. The plan is for the model to analyse a series of patient parameters and, for example, inform the doctor about which types of treatments have worked for other individuals with very similar conditions.

A living tool for living hearts

Initially, Cardiomentor will be a training tool: it will help general practitioners stay updated and make decisions. Moreover, it is planned that the model will receive constant updates, incorporating new information as it gets published. Once the assistant is ready, the plan is to incorporate real-life cases. This will be achieved by introducing data from the medical records of patients who have suffered from heart diseases into the model.

Gigantic amounts of data will be needed to personalise treatments. To do this, variables such as family situation, diet, and type of work will be recorded to create something similar to digital twins, according to the app developers. Before becoming operational, the tool will be clinically validated by the Spanish Society of Cardiology. The app will help deal with Spain’s main cause of death: heart disease, which, along with heart failure, led to 46,688 deaths in 2024, according to the National Institute of Statistics of Spain.

The Spanish foundational model

The family of generative AI models, Alia, presented by the country’s government earlier this year, does not aim to compete with the big players in the sector, such as Llama, ChatGPT, Perplexity, Gemini, or DeepSeek, but rather to provide a tool that works well in Spanish and 35 European languages.

Alia is an open-source project, meaning that anyone, whether an individual or a company, can download and modify it. In addition, the model has been trained with official public documentation, ranging from historical records of the BOE (Official State Gazette) and the Official Gazette of the Commercial Registry to parliamentary interventions and judicial rulings. The government has allocated 2.2 million euros to develop the algorithm and another eight million for the necessary infrastructure provided by BSC and for configuring the databases. Work is already underway on designing new use cases beyond Cardiomentor, and an internal pilot application for the Tax Agency is aimed at streamlining bureaucratic procedures.