CURRENT EDITION

MLOps, Continuous Learning, and Resource Management in the Edge-Cloud Continuum

May 28, 2025 - 9:00 - 10:30

Room 1 (Σ010)

This workshop will explore innovative techniques for cognitive management and orchestration in machine learning workflows and Edge-Cloud resources. It will showcase the latest advancements from several European research and innovation projects, including CODECO, COGNIFOG, MLSysOps and ENACT. These projects develop AI-driven edge-cloud resource management strategies to optimize performance, scalability, security and energy efficiency. In addition, the M4ESTRO and S-X-AIPI projects will be featured for their contributions to applied automation and the development of robust MLOps pipelines.

Participants will gain insights into:

· The critical challenges in ML lifecycle management, highlighting issues such as data drift, model degradation, and concept drift, and their impact on long-term model reliability.

· The importance of Human-in-the-Loop (HITL) learning in refining ML models, focusing on active learning workflows and expert annotation strategies.

· MLOps pipelines: covering the end-to-end ML lifecycle, from data ingestion and preprocessing to model training, validation, deployment, and monitoring. Participants will explore workflow automation strategies, including CI/CD integration, version control, and orchestration tools, to enable scalable and reproducible machine learning operations.

· Advanced AI and ML-driven automation techniques for resource allocation, workload distribution, and real-time system adaptation across the edge-cloud continuum.

· Comparative analysis of centralized, decentralized, and hybrid orchestration strategies for managing distributed AI/ML workloads.

· Key challenges in orchestrating heterogeneous edge-cloud environments, including network reliability, data privacy, multi-tenancy, and scalability.

· Approaches for addressing conflicting objectives and operational trade-offs in resource management and orchestration.

· Best practices, methodologies, and lessons learned from leading EU-funded research projects.

· Strategies for integrating MLOps workflows with resource management to enable continuous learning, adaptive model deployment, and automated pipeline management.

· Industrial case studies and real-world use cases demonstrating the practical value of AI and MLOps-enabled resource management.

· Opportunities for collaboration and knowledge exchange among researchers, industry stakeholders, and policymakers to drive future advancements and implementations in edge-cloud AI systems.

Through presentations and a discussion panel, the session will provide a platform for researchers, industry practitioners, and policymakers to exchange knowledge and explore future directions for MLOps and resource management in the Edge-Cloud continuum.

Speakers/panelists:

Damir Filipovic, Alliance for AI, IoT and Edge Continuum Innovation

Christos Antonopoulos, University of Thessaly, MLSysOps Project,

Grigoris Antonopoulos, Netcompany, COGNIFOG Project.

Usman Wajid, Information Catalyst, ENACT Project.

Athanasia Sakavara, Core Innovation.

Ioannis Meintanis, Core Innovation

Asier Arteaga Ayarza, Sidenor

Dorine Mantzakou, Netcompany-Intrasoft

John Soldatos, Netcompany-Intrasoft – CODECO Project

Projects contributing to the workshop/session (if any)

CODECO, funded under the Horizon Europe program, Grant Agreement No: 101092696.
s-X-AIPI, funded under the Horizon Europe program, Grant Agreement No: 101058715.
M4estro, funded under the Horizon Europe program, Grant Agreement No: 101138506
COGNIFOG, funded under the Horizon Europe program, Grant Agreement No: 101092968.
MLSysOps, funded under the Horizon Europe program, Grant Agreement No: 101092912.
ENACT, funded under the Horizon Europe program, Grant Agreement No: 101135423.
COGNIT, funded under the Horizon Europe program, Grant Agreement No: 101092711.
DECICE, funded under the Horizon Europe program, Grant Agreement No: 101092582.

Audience engagement (explain the format of the session/workshop)

Speakers Presentations (60 min)

10’ “Advancing AI-Driven Innovation Across the Edge-Cloud-IoT Continuum”, Damir Filipovic, Alliance for AI, IoT and Edge Continuum Innovation.
10’ “Cloud-Edge Continuum Cross-Layer Orchestration and Resource Management: The CODECO Approach”, John Soldatos, Netcompany-Intrasoft
10’ “Cognitive Fog Computing for Efficient Resource Management in the Edge-Cloud Continuum””, Grigoris Antonopoulos, Netcompany
10’ “MLOps-based Autonomic System Management in the Edge-Cloud Continuum”, Christos Antonopoulos, University of Thessaly.
10’ “Adaptive Intelligence and Dynamic Resource Management in the Cognitive Edge-Cloud Continuum: ENACT Project Innovations”, Usman Wajid, Information Catalyst.
10’ “Why Models Fail: Leveraging MLOPs and active learning for resilient and adaptive AI”, Athanasia Sakavara, Core IC

Panel Discussion & Audience Q&A (30 min)

Moderated panel featuring expert insights on building resilient machine learning pipelines and orchestrating AI workloads across heterogeneous environments. The discussion will cover centralized vs. decentralized orchestration strategies, the role of AI and MLOps in edge-to-cloud resource management, and the key challenges in real-world implementation. Each panelist will answer 1-2 Questions (20’).
Open Q&A where the audience can engage directly with panelists. Use of Polls or quick feedback via interactive tools (e.g., Mentimeter, Slido) to gather input from attendees on key challenges in Edge-Cloud orchestration (10’)