Edge computing offers key advantages over traditional cloud-based solutions, particularly in areas such as privacy, safety, end-to-end latencies, cost, and energy efficiency. The trend toward Edge AI, which involves running artificial intelligence workloads closer to sensors and actuators on resource-constrained devices, is driving innovation. According to a market report, the Edge AI market is expected to grow significantly from USD 17.46 billion in 2023 to nearly USD 62.93 billion by 2030. EdgeAI has been studied already for quite some time and there exist mature flows and edge compute platforms to integrate AI components. Yet, these mainly focus on the deployment step (initial setup) and not the complete lifecycle of an edge solution and they do not consider additional requirements that arise in industrial and safety critical applications. A holistic view on the AI lifecycle is required to enable core Austrian industries in fields such as transportation, manufacturing or renewable energy to deploy EdgeAI solutions in their systems.  

To address these challenges, the DOMINO project aims to create an advanced MLOps lifecycle for Edge AI, inspired by the DevOps cycle used in software development. This approach focuses on the entire lifecycle of Edge AI systems, from initial development and deployment to monitoring, evaluation and updating. In the initial deployment phase, DOMINO will focus on also addressing requirements for industrial and safety critical applications such as integration in IT environments of automation systems and safety norms derived from the EU AI Act. Even if the system fullfills all requirements in the initial deployment, operation conditions and AI systems may drift or degrade over time. Hence, DOMINO will also investigate methods to monitor the functionality of the AI system once it is deployed in the field. As ML Experts might be involved in the design of the AI components but not in their direct operation in the field, DOMINO will investigate methods to use non-ai-expert operator and user feedback for evaluating Edge AI components. Based on this feedback and the monitoring data, DOMINO will also investigate how the system can be improved by data collected in the field without “breaking the system”, e.g., due forgetting previously learned information.  

For this, DOMINO brings together industry and research partners to develop advanced solutions and will demonstrate the innovative methods in three use cases: Wood production, rail transportation and renewable energy. Project goals are to lower the barriers to deploying AI on edge devices, ensure compliance with EU regulations for AI in safety-critical systems and improve system monitoring in real-world environments. Due to focusing in the challenges which arise from bringing AI systems “into the wild”, we expect that DOMINO will significantly improve the competitiveness of Austrian industry, that are often hidden champions in domains that need to address industrial and safety constraints. Additionally, the selected use cases are key areas for the green transformation to a sustainable society, hence, enabling them to improve the products in these fields with EdgeAI components will support this transformation.