Loading...
Advancing domain-specific AI/ML integrations: the role of low-code platforms for responsible AI
Date
2025
Abstract
This thesis proposes a unified vision and approach for Responsible AI that is based on the four pillars of reusability, interpretability,
accessibility, and interoperability. This approach is validated through real-world applications in the domains of healthcare, Internet of Things
(IoT)-based Cyber-Physical Systems (CPS), and domain-specific work-flows for domain experts. This thesis aims to integrate the cutting-edge
research ideas from the fields of Artificial Intelligence (AI), Machine Learning (ML), software development, CPS, and IoT to address the
technical and human-centric challenges for the development, deployment and use of AI/ML-based solutions. By merging the concepts from
Low-Code/No-Code (LC/NC) frameworks, domain-specific languages, and model-driven engineering, this thesis demonstrates that AI/ML-based solutions can be made more approachable for domain experts and end users. In this context, modular LC/NC platforms like Pyrus, ADD-Lib, D ime, C inco de Bio allowed for the rapid prototype development and iterative experimentation for various on-device to cloud deployments. At the same time, IoT hardware solutions, like THINGY : 53 and Raspberry Pi-based expansions, provide case studies in practical real-world environments with resource constraints. This system-level approach in combination with rapid feedback loops with stakeholders not only helps to bridge the gaps between domain experts and development of domain-specific solutions but also highlights the demand for accessible and reliable edge-based analytics. The viability of the proposed solutions was validated and confirmed in collaborations with several industry and academic institutes in Ireland and Europe. The research outputs from these collaborations showed that the combination of user-centric design, sustainable software development methodologies, and flexible edge-based deployments can produce outcomes that align with the practical constraints and requirements of the domain experts and end users. The successful implementation and use of AI/ML-based solutions to address the challenges in various domains requires more than just powerful models and algorithms, and actually depends on the whole software-hardware ecosystem that allows the domain experts to themselves extend existing solutions for new challenges without compromising on transparency or safety.
Supervisor
Margaria, Tiziana
Description
Publisher
University of Limerick
Citation
Files
Loading...
Singh_2025_Advancing.pdf
Adobe PDF, 11.35 MB
Funding code
Funding Information
Sustainable Development Goals
Type
Thesis
Rights
http://creativecommons.org/licenses/by-nc-sa/4.0/