kNOWLEDGE HUB
Unlocking Wisdom, One File at a Time
Newsletter #1
Our 1st newsletter edition, introducing you to the AI-DAPT concept. May 2024
Newsletter #2
Our 2nd newsletter edition, introducing our four real-life demonstrators and research agenda. January 2025
Newsletter #3
About platform’s technical implementation, Reference Architecture, and upcoming participation at ADRF25.
Newsletter #4
About baseline platform services, data-centric foundations, and hybrid Science–AI models for reproducible and adaptive AI pipelines..
Leaflet #1
Leaflet about AI-DAPT, presenting its main pilots, research agenda and concept.
Effector: A Python package for regional explanations
Effector is a Python library for analyzing regional feature effects, helping interpret models by identifying where feature impacts vary across subgroups.
Fast and Accurate Regional Effect Plots for Automated Tabular Data Analysis
r-RHALE is a fast explainability method that shows how feature effects vary across subgroups in tabular data, even with large datasets and complex models.
Big Data Visualisation in the Maritime Industry
VesselAI reviews big data tools for maritime needs and shows how Apache Superset can visualize AIS data, routes, and weather to support decision-making.
Utilizing large-scale human mobility data to identify determinants of physical activity
Using mobility data from 20M+ people, this study reveals how visits to places like cafes, parks, and grocery stores relate to fitness habits.
Bridging Data and AIOps for Future AI Advancements with Human-in-the-loop.
The AI-DAPT Concept
AI-DAPT uses synthetic data and hybrid ML to build trustworthy AI for healthcare, robotics, energy, and manufacturing.
Data Ingestion and Harmonisation for the Maritime Domain
Maritime data is diverse and complex. A new service uses NLP to unify raw maritime data into a common format, enabling better AI and machine learning applications—no coding required.
D1.1 Automated AI Pipeline End User Needs and Scientific Technology Radar
D1.1 defines AI-DAPT’s user needs, ethical principles, and tech focus areas—like XAI, synthetic data, and hybrid models—based on input from its four real-world pilots.
D1.2 Automated AI Pipeline Design & Technical Requirements
Outlines AI-DAPT’s technical needs, user roles, and system workflows, forming the foundation for platform architecture.
D6.1 Dissemination Communication Engagement Exploitation
Details AI-DAPT’s communication and engagement plan, including key channels, early outreach, and target audiences.
D7.1 Project Management Handbook
The handbook guides AI-DAPT’s management, detailing roles, timelines, and processes to develop a trustworthy, automated AI framework combining data, models, and human input.
D7.2 ARGOS DMP
The AI-DAPT DMP defines how project data is managed using FAIR principles, ensuring security, accessibility, and ethical handling across health, robotics, and energy demonstrators.
D7.2 Data Management Plan
The DMP outlines how AI-DAPT manages data ethically and securely, following FAIR principles, across its health, robotics, energy, and manufacturing use cases.
Factsheet
Flyer about AI-DAPT's factsheet, presenting its Horizon Europe RIA Project, call, grant ID and duration.
PROMIS: A Post-Processing Framework for Mitigating Spatial Bias
A new post-processing method reduces spatial bias in machine learning models while preserving accuracy, delivering fairer outcomes without access to the original training data.
Hybrid Intelligence: The Fusion of Science-Based and Machine Learning Models
Hybrid models combine science-based knowledge with machine learning to improve prediction accuracy and interpretability.
Roadmaps and Agendas for Research and Innovation in Artificial Intelligence, Data, and Robotics
AI, data, and robotics research focuses on innovation, ethics, and explainable systems to build trustworthy and competitive technologies.
GLANCE: Global Actions in a Nutshell for Counterfactual Explainability
GLANCE provides simple, low-cost counterfactual actions that explain ML decisions while maximizing impact and interpretability.
Socioeconomic and demographic disparities in fitness center visits
Socioeconomic factors like income and education strongly predict fitness center usage, revealing disparities in physical activity across communities.
Fast and Robust Simulation-Based Inference With Optimization Monte Carlo
A gradient-based simulation inference method enables faster Bayesian posterior estimation in complex stochastic models with reduced runtime and fewer simulations.
A Unified Data Staging Area for Integrated Analytics
A unified data staging system abstracts heterogeneous data sources, enabling seamless, efficient access and integration in analytics pipelines.
Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects
CALMs improve GAMs by adding region-specific effects that capture interactions while keeping models interpretable and accurate.
The Human Oversight approaches at the forefront of responsible and trustworthy AI, from data-centric adaptive AI-Ops pipelines to Multi-Agent Systems
Human oversight in AI (HITL/HOTL) improves accountability and fairness but introduces scalability, bias, and workload challenges.
Hybrid Model Engineering: A Residual Learning Approach for Modular AI Pipelines
A modular framework enables reproducible hybrid modeling by combining physics-based models with machine learning in structured pipelines.
A Microservice-based Architecture for Reproducible AI Pipelines
A unified AI pipeline architecture integrates data, models, and governance into a single lifecycle system to improve reproducibility and trustworthiness.
Bridging Complexity and Usability: The DAVE Visual Execution Environment for AI/ML Pipelines
A visual execution environment makes AI/ML pipelines easier to monitor, debug, and understand through task-level observability and simplified interfaces.
The Convergence of Data-Centric Engineering and Adaptive AI Operations: A Comprehensive Framework for Trustworthy Hybrid Systems
Data-centric AI combines synthetic data, observability, and adaptive learning to build more reliable and trustworthy AI systems.
Newsletter #1
Our 1st newsletter edition, introducing you to the AI-DAPT concept. May 2024
Newsletter #2
Our 2nd newsletter edition, introducing our four real-life demonstrators and research agenda. January 2025
Newsletter #3
About platform’s technical implementation, Reference Architecture, and upcoming participation at ADRF25.
Newsletter #4
About baseline platform services, data-centric foundations, and hybrid Science–AI models for reproducible and adaptive AI pipelines..
D1.1 Automated AI Pipeline End User Needs and Scientific Technology Radar
D1.1 defines AI-DAPT’s user needs, ethical principles, and tech focus areas—like XAI, synthetic data, and hybrid models—based on input from its four real-world pilots.
D1.2 Automated AI Pipeline Design & Technical Requirements
Outlines AI-DAPT’s technical needs, user roles, and system workflows, forming the foundation for platform architecture.
D6.1 Dissemination Communication Engagement Exploitation
Details AI-DAPT’s communication and engagement plan, including key channels, early outreach, and target audiences.
D7.1 Project Management Handbook
The handbook guides AI-DAPT’s management, detailing roles, timelines, and processes to develop a trustworthy, automated AI framework combining data, models, and human input.
D7.2 ARGOS DMP
The AI-DAPT DMP defines how project data is managed using FAIR principles, ensuring security, accessibility, and ethical handling across health, robotics, and energy demonstrators.
D7.2 Data Management Plan
The DMP outlines how AI-DAPT manages data ethically and securely, following FAIR principles, across its health, robotics, energy, and manufacturing use cases.
Effector: A Python package for regional explanations
Effector is a Python library for analyzing regional feature effects, helping interpret models by identifying where feature impacts vary across subgroups.
Fast and Accurate Regional Effect Plots for Automated Tabular Data Analysis
r-RHALE is a fast explainability method that shows how feature effects vary across subgroups in tabular data, even with large datasets and complex models.
Big Data Visualisation in the Maritime Industry
VesselAI reviews big data tools for maritime needs and shows how Apache Superset can visualize AIS data, routes, and weather to support decision-making.
Utilizing large-scale human mobility data to identify determinants of physical activity
Using mobility data from 20M+ people, this study reveals how visits to places like cafes, parks, and grocery stores relate to fitness habits.
Bridging Data and AIOps for Future AI Advancements with Human-in-the-loop.
The AI-DAPT Concept
AI-DAPT uses synthetic data and hybrid ML to build trustworthy AI for healthcare, robotics, energy, and manufacturing.
Data Ingestion and Harmonisation for the Maritime Domain
Maritime data is diverse and complex. A new service uses NLP to unify raw maritime data into a common format, enabling better AI and machine learning applications—no coding required.
PROMIS: A Post-Processing Framework for Mitigating Spatial Bias
A new post-processing method reduces spatial bias in machine learning models while preserving accuracy, delivering fairer outcomes without access to the original training data.
Hybrid Intelligence: The Fusion of Science-Based and Machine Learning Models
Hybrid models combine science-based knowledge with machine learning to improve prediction accuracy and interpretability.
Roadmaps and Agendas for Research and Innovation in Artificial Intelligence, Data, and Robotics
AI, data, and robotics research focuses on innovation, ethics, and explainable systems to build trustworthy and competitive technologies.
GLANCE: Global Actions in a Nutshell for Counterfactual Explainability
GLANCE provides simple, low-cost counterfactual actions that explain ML decisions while maximizing impact and interpretability.
Socioeconomic and demographic disparities in fitness center visits
Socioeconomic factors like income and education strongly predict fitness center usage, revealing disparities in physical activity across communities.
Fast and Robust Simulation-Based Inference With Optimization Monte Carlo
A gradient-based simulation inference method enables faster Bayesian posterior estimation in complex stochastic models with reduced runtime and fewer simulations.
A Unified Data Staging Area for Integrated Analytics
A unified data staging system abstracts heterogeneous data sources, enabling seamless, efficient access and integration in analytics pipelines.
Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects
CALMs improve GAMs by adding region-specific effects that capture interactions while keeping models interpretable and accurate.
The Human Oversight approaches at the forefront of responsible and trustworthy AI, from data-centric adaptive AI-Ops pipelines to Multi-Agent Systems
Human oversight in AI (HITL/HOTL) improves accountability and fairness but introduces scalability, bias, and workload challenges.
Hybrid Model Engineering: A Residual Learning Approach for Modular AI Pipelines
A modular framework enables reproducible hybrid modeling by combining physics-based models with machine learning in structured pipelines.
A Microservice-based Architecture for Reproducible AI Pipelines
A unified AI pipeline architecture integrates data, models, and governance into a single lifecycle system to improve reproducibility and trustworthiness.
Bridging Complexity and Usability: The DAVE Visual Execution Environment for AI/ML Pipelines
A visual execution environment makes AI/ML pipelines easier to monitor, debug, and understand through task-level observability and simplified interfaces.
The Convergence of Data-Centric Engineering and Adaptive AI Operations: A Comprehensive Framework for Trustworthy Hybrid Systems
Data-centric AI combines synthetic data, observability, and adaptive learning to build more reliable and trustworthy AI systems.
Leaflet #1
Leaflet about AI-DAPT, presenting its main pilots, research agenda and concept.
Factsheet
Flyer about AI-DAPT's factsheet, presenting its Horizon Europe RIA Project, call, grant ID and duration.