LLM4ALZ-HPC: Advanced Computing for a Local-First LLM and Multi-Agent Environment for Alzheimer’s Disease Decision Support

LLM4ALZ-HPC aims to develop and evaluate a local-first artificial intelligence environment for Alzheimer’s disease research that integrates open-weight large language models, Retrieval-Augmented Generation (RAG), and multi-agent systems. Using the Deucalion supercomputing infrastructure, the project will benchmark model performance, efficiency, reliability, and traceability to support research on diagnostic decision-making, longitudinal risk assessment, and cognitive-decline prediction. The project will deliver a research proof of concept and support two Master’s dissertations, establishing a foundation for future clinical validation

Keywords: Alzheimer’s Disease, Artificial Intelligence, High-Performance Computing, Large Language Models, Multi-Agent Systems, Retrieval-Augmented Generation

Principal Investigator: Rafael Pinheiro  (rafael.f.pinheiro@ulo.pt)

Research Team:

• ciTechCare: Rui Fonseca-Pinto (rui.pinto@ulo.pt); Remy Cardoso
• Escola de Tecnologia e Gestão  – Universidade de Leiria e Oeste (ULO): Conceição Nogueira; Hugo Rafael Godinho Lopes; Maria João da Silva Alves

Duration: 12 months (2026 – 2027)

Institutional Partners and Others

• DEUCALION

Reference: 2026.08290.CPCA.A1

Funding: Portuguese Foundation for Science and Technology