RAPTURE

As part of Fernando’s EngD research project, a significant contribution was made to the field of HPC through the development of Job Prophet (JP), an innovative AI-powered tool designed to enhance memory allocation. Addressing the complex challenge of efficient memory management in large-scale HPC infrastructures, the EngD candidate applied Action Design Research  and CRISP-DM to identify and implement best practices and intelligent automation strategies. By leveraging real-world HPC workload data and embedding continuous training, inference, and monitoring through robust CI/CD pipelines, JP demonstrated the ability to dynamically adapt to changing workloads while improving operational efficiency. The project not only delivered measurable gains in resource utilisation and job performance but also showcased the value of AI-driven decision-making in complex, data-intensive computing environments. This successful case exemplifies how rigorous research and applied engineering can converge to solve real-world challenges in HPC system optimisation.

In Pradheepan’s EngD project, a large language model (LLM)-based chatbot was developed to enhance user support within the NXP Design Infrastructure (NxDI), a critical platform used for managing semiconductor design simulations and workflows. The chatbot was created to intelligently handle high volumes of repetitive and routine support queries, which traditionally consumed significant time and resources from the engineering support team. By automating these interactions, the chatbot enabled faster response times and improved efficiency, allowing support staff to focus on complex, high-value tasks. The project employed cutting-edge natural language processing techniques to ensure accurate, context-aware responses tailored to the unique requirements of semiconductor design. Through continuous refinement based on real user feedback and integration within NxDI’s operational environment, the chatbot evolved into a reliable digital assistant that significantly improved the user experience. This successful implementation illustrates how LLM technologies can be harnessed within specialised engineering domains to drive operational excellence and innovation

Stefano Fossati’s PhD project focuses on improving HPC infrastructure observability and management through advanced visualisation techniques. The research aims to align architectural intent with operational reality, enhancing monitoring capabilities and increasing infrastructure efficiency. By providing visibility on resource utilisation, we help organisations to improve efficiency and streamline operations, while enabling deeper insights into system behaviour and supporting improved management and optimisation. The project also emphasises failure analysis of HPC systems. By delivering robust mechanisms for performance inspection and issue diagnosis, it aims to improve the reliability and resilience of complex infrastructures.

Marco Tonnarelli’s PhD project is situated within the evolving landscape of modern, data product-oriented, decentralized big data architectures. The objective of his project is to develop a framework for automated policy-as-code enforcement within federated, computationally governed organizations. The research is organized across three interconnected layers: governance, architecture, and code. By identifying and analysing the key organizational challenges that arise during transitions to such architectures, Marco aims to explore architectural patterns from cutting-edge metadata management, an essential component in these systems, and ultimately design an automated framework for federated policy enforcement, leveraging Natural Language Processing (NLP) technologies. Methodologically, the project combines primary and secondary approaches from empirical software engineering. Due to the limited academic coverage of this emerging topic, industry insights are essential. The research currently includes a multivocal systematic literature review and a qualitative study, with plans to incorporate mining studies, ethnographic methods, case studies, and action research as it progresses.