Imagine trying to bake a cake but the ingredients are messy and inconsistent; the cake will fail regardless of the recipe. This project builds a smart system that automatically cleans and prepares the 'ingredients' (data) while letting a human expert double-check the quality. It also blends traditional scientific laws with AI to make sure the results are predictable and safe.
AI models often fail because the data feeding them is poor or biased, and fixing this manually is too slow. Additionally, pure AI models can be unreliable because they ignore established scientific laws.
An automated AI-Ops system for managing the lifecycle of data and AI pipelines, including tools for automated cleaning, synthetic data generation, and hybrid science-AI model coupling.
If you are a medical diagnostic software provider dealing with inconsistent patient data—this project developed an automated data-centric pipeline that improves reliability and fairness. It ensures high-quality data is used to train AI, reducing the time it takes to get clinical insights.
If you are an industrial robot manufacturer dealing with unpredictable AI behavior in physical environments—this project developed a hybrid system coupling science-based models with AI. This ensures the robot follows physical laws while learning from data, increasing safety and reliability.
If you are a smart grid operator dealing with massive, noisy sensor data—this project developed AI-driven automation for data cleaning and observability. This allows for continuous, dynamic improvements to energy prediction models without manual data scrubbing.
Based on available project data, specific pricing or licensing costs are not provided as this is a research project funded by the EU.
Yes, the project specifically aims to create scalable data-AI pipelines and will validate results across 4 representative industries including Health, Robotics, Energy, and Manufacturing.
Based on available project data, the specific IP terms are not listed, but the results will be integrated into both open source and commercial AI solutions currently on the market.
The system is designed to be integrated into existing AI solutions, whether they are commercial or open source, to improve their data handling and model reliability.
The project period runs from 2024-01-01 to 2027-06-30, suggesting that fully validated results will be available by mid-2027.
The consortium is heavily weighted toward commercial application, with 8 industry partners (including 10 SMEs across the group) representing a 44% industry ratio. This suggests a strong focus on market viability, supported by 4 universities and 5 research centers across 7 European countries.
Contact ATHINA-EREVNITIKO KENTRO KAINOTOMIAS STIS TECHNOLOGIES TIS PLIROFORION
Contact us to connect with the AI-DAPT consortium for early adoption of automated AI-Ops pipelines.
