AggRaDa is a framework designed to automate the spatiotemporal aggregation of heterogeneous raw data from multiple sources. Its purpose is to automatically identify granularities that maximize the consistency and representativeness of the data, making them more useful for analysis and decision-making. The modular microservices-based architecture integrates indexing and manipulation of spatial and temporal data, enabling efficient exploration of multiple scales and quantitative evaluation of the results.

Automatic spatiotemporal aggregation of multi-source heterogeneous raw data at multiple granularities.

large_aggrada_thumbnail.webp

medium_aggrada_thumbnail.webp

small_aggrada_thumbnail.webp

thumbnail_aggrada_thumbnail.webp

aggrada_thumbnail.webp

AggRaDa - Aggregation of Raw Data

Constructing elementary cell of Cellular Neural Networks (CelNN) in an IBM physical quantum computer. An elementary cell works as a Full Adder for digital computers. CelNN are among a dozen candidates for quantum NN.

CellNN has high locality and cells that can use low amounts of qubits, which may enable them in a Noise-Intermediate State Quantum (NISQ) device. 

medium_Screenshot_2024_09_12_at_09_51_50_21f02e1b9a.png

small_Screenshot_2024_09_12_at_09_51_50_21f02e1b9a.png

thumbnail_Screenshot_2024_09_12_at_09_51_50_21f02e1b9a.png

Screenshot_2024_09_12_at_09_51_50_21f02e1b9a.png

CelNN in a quantum device

Innovative tool that uses FAIR quality metrics, Openness, and data maturity to automatically assess the quality and transparency of databases through the attribution of scores and detailed reports.

Automatically evaluate databases quality and transparency.

large_dbscore_thumb.webp

medium_dbscore_thumb.webp

small_dbscore_thumb.webp

thumbnail_dbscore_thumb.webp

dbscore_thumb.webp

DBScore

EEG processing does not follow a universally accepted workflow. Different research communities adopt distinct approaches influenced by experimental goals, participant populations, available equipment, and target metrics.

As a consequence, researchers and students often need to navigate a fragmented literature to understand which methods exist, when they should be applied, and which parameters are critical in each context.

This project aims to reduce this fragmentation through a systematic mapping of methodological options currently used in EEG research.

Interactive knowledge platform that maps the methodological landscape of EEG signal processing, connecting techniques, parameters, and analytical pathways across the entire processing pipeline.

large_eeg methodology maps.png

medium_eeg methodology maps.png

small_eeg methodology maps.png

thumbnail_eeg methodology maps.png

eeg methodology maps.png

EEG Methodology Map

This project provides a structured validation framework to assess whether a large language model (LLM) used in healthcare can safely classify a case as emergency, urgent, routine, or self-care, identify warning signs, and remain reliable under realistic stress conditions such as noisy input, lay language, conflicting information, and manipulation attempts. Its purpose is to reduce risk before deployment by generating objective evidence for go/no-go decisions, procurement, governance, and audit. This need is consistent with the World Health Organization’s guidance on the governance of generative AI in health.

Projects

Outreach of Center for Artificial Intelligence in Health Management

AggRaDa - Aggregation of Raw Data

CelNN in a quantum device

DBScore

EEG Methodology Map

MedEval Triage