CORE Group at Data Week 2025
Author: Konstantina Tsioli, Athanasia Sakavara
17th June 2025
This year’s Data Week, organised by the Big Data Value Association (BDVA) in collaboration with the CORE Innovation Centre, took place in Athens last month.
This year’s event focused on the fundamental elements of Data Value creation. Artificial Intelligence (AI) serves as a pivotal tool in unlocking data value through various means, hence the event’s name, “Odyssey of AI: Navigating the Data Seas”.
Our team was invited to participate in two sessions.
Manufacturing Data Spaces and the European approach to AI
During this BDVA session, we discussed how AI on Demand (AIoD) and European Digital Infrastructure Consortia (EDICs) can accelerate manufacturing-oriented verticals, aligning with the product lifecycle (design, production, supply chain, and end-of-life). The Digital Transformation department from CORE INNOVATION, highlighted the work done within EU initiatives to pilot AI-driven manufacturing solutions—such as modular and adaptive production (MODUL4R) and dynamic supply-chain optimization in servitised manufacturing frameworks (M4ESTRO).
Despite Europe’s booming AI ecosystem, only 13.5% of companies have adopted AI, underscoring the need for structured frameworks like AIoD (offering ready-made tools for rapid prototyping, generative design, and quality control) and EDICs (providing scalable infrastructure like material science LLMs, digital twin platforms, and circular economy data spaces).
The session emphasised AI Continent’s five pillars (computing, data, skills, regulation, and strategic adoption), positioning AIoD as the front-end for industry-ready tools (e.g., federated datasets for supply chain optimization) and EDICs as the backbone for heavy-duty infrastructure (e.g., cross-border data spaces). TEFs and EDIHs bridge adoption gaps, but open questions remain on open-source accessibility and compliance efficiency. The goal? A cohesive ecosystem where stakeholders—from SMEs to R&D—can leverage standardized, compliant AI solutions to fast-track manufacturing innovation.
MLOps, Continuous Learning, and Resource Management in the Edge-Cloud Continuum
For this session, our contributions included shaping the agenda, participating in a panel discussion, and delivering the presentation “Why Models Fail: Leveraging MLOps and Active Learning for Resilient and Adaptive AI.”
In our talk, we addressed key challenges in the machine learning lifecycle, focussing on data drift and concept drift, which often lead to model degradation—and emphasized the importance of Human-in-the-Loop (HITL) methods to build more trustworthy and adaptive AI systems. We highlighted the value of modern MLOps practices such as CI/CD pipelines, automation, versioning, monitoring, and active learning integration, all of which are essential for maintaining high-performing AI in dynamic, real-world environments.
Going from theory to practice, we presented two solutions we developed in ongoing R&D projects that exemplify scalable, automated, and reliable MLOps systems: in S-X-AIPI, we developed ADAPT, an intelligent anomaly detection system; and in M4ESTRO, we introduced a multi-agent reinforcement learning framework that delivers actionable insights for optimizing adaptive value chains in manufacturing.
