An Agentic AI Assistant for IPv6 Planning, Configuration & Operations (GENIE)
Call Identifier: OCF-OC1-Challenge2 (download call pdf)
Objectives
The project addresses the OCF challenge of using AI agents to support IPv6 education, deployment and operations, where relevant knowledge is spread across hundreds of IETF RFCs and vendor manuals. It will evaluate how current LLMs perform beyond question answering, including IPv6 address planning, lab testing, deployment guidance and operational support, and will develop GENIE6 as an evidence-grounded assistant for network operators. GENIE6 combines standards-aware retrieval, citation-backed reasoning, operator-oriented lab scenarios, configuration planning, automated error identification and validation, and methods for combining multiple LLM outputs to improve reliability. The expected outcome is a reproducible prototype and practical guidance for using AI to help IPv6 administrators produce traceable plans, lab configurations, validation checks and operational recommendations.
Principal Investigator
Prof. Peyman Teymoori is Professor of Computer Science at the University of South-Eastern Norway (USN), Department of Business, Strategy and Political Sciences, USN School of Business, Campus Drammen. His research focuses on network modelling, optimisation and performance analysis, next-generation communication systems, and the use of AI/ML techniques for resilient and secure networking.
Sponsor Technical Expert: XiPeng Xiao
EPD-Aware Prefill-Centric Traffic Slicing for Multimodal LLM Inference (SLICE-EPD)
Call Identifier: OCF-OC1-Challenge4 (download call pdf)
Objectives
The project addresses the OCF challenge of using AI agents to support IPv6 education, deployment and operations, where relevant knowledge is spread across hundreds of IETF RFCs and vendor manuals. It will evaluate how current LLMs perform beyond question answering, including IPv6 address planning, lab testing, deployment guidance and operational support, and will develop GENIE6 as an evidence-grounded assistant for network operators. GENIE6 combines standards-aware retrieval, citation-backed reasoning, operator-oriented lab scenarios, configuration planning, automated error identification and validation, and methods for combining multiple LLM outputs to improve reliability. The expected outcome is a reproducible prototype and practical guidance for using AI to help IPv6 administrators produce traceable plans, lab configurations, validation checks and operational recommendations.
Principal Investigator
Dr. Mohammad Shojafar is an Associate Professor in Network Security at the Institute for Communication Systems (ICS) and the 6G Innovation Centre (6GIC), University of Surrey, UK. He is a Senior Member of IEEE, ACM Distinguished Speaker, Intel Innovator, and Fellow of the Higher Education Academy, with over 15 years of combined academic and industrial experience across AI-native networking, cybersecurity, Open RAN, cloud/edge computing, and intelligent 5G/6G systems. His research focuses on AI-driven network optimisation, trustworthy and explainable AI for communication systems, network slicing, autonomous O-RAN control, and secure large-scale distributed infrastructures.
Sponsor Technical Expert: Bruno Ricardo da Cunha Magalhaes
INtegrated Security via Intelligent Graph and Heuristic Tracking (INSIGHT)
Call Identifier: OCF-OC1-Challenge5 (download call pdf)
Objectives
The INSIGHT project aims to improve cyber threat detection by combining advanced artificial intelligence with a structured understanding of how data flows through digital systems. Rather than relying solely on predefined rules or historical patterns, it represents systems as interconnected networks, enabling the tracing of data provenance and relationships across components. By integrating machine learning with interpretable reasoning techniques, the approach can adapt to emerging and evolving threats while providing explanations for its decisions. The result is a more adaptive and transparent approach to cyber security that helps organisations detect issues earlier and more efficiently, understand them more clearly, and make more informed and effective responses.
Principal Investigator
Dr. Martin Barrere is a Lecturer (Assistant Professor) in Digital Resilience at the Surrey Centre for Cyber Security, Computer Science Research Centre, University of Surrey, UK, and an Honorary Research Fellow at the Institute for Security Science and Technology at Imperial College London, UK. His current work focuses on cyber security for critical national infrastructure. In 2020, he was awarded a GCHQ Research Fellowship for National Resilience. He received his PhD degree in Computer Science from the University of Lorraine / INRIA, France, in 2014. His main research interests include cyber security, computer networks, cyber-physical systems, artificial intelligence, attack graphs, security metrics, risk analysis, autonomic computing, and forensic investigations.
Sponsor Technical Expert: Zied Ben Houidi
Lightweight In-Network Monitoring with Efficient End-Host Recovery (LIMER)
Call Identifier: OCF-OC1-Challenge3 (download pdf)
Objectives
The LIMER project addresses the loss of compute and time caused by hardware and communication faults in the GPU clusters used to train large-scale AI models, where a single faulty link or device can stall an entire distributed job and force a restart from the last checkpoint. The project moves the first line of fault detection inside the network: the top-of-rack switch runs lightweight, high-recall detection within tight memory limits and signals the affected GPU server, which refines that signal using an interpretable fuzzy-rule classifier and calibrated uncertainty estimation, all without modifying standard training stacks. By combining in-network detection with graceful recovery that reconfigures GPU communication rather than restarting from a checkpoint, the project aims to reduce detection latency, retain training throughput during faults, and avoid unnecessary job restarts. Expected outcomes include a validated in-network detection-and-recovery method, reusable fault-injection and evaluation tooling, and deployment guidance for resilient AI-native GPU-cluster infrastructures.
Principal Investigator
Dr. Xilu Wang is a Lecturer and Surrey Future Fellow at the Computer Science Research Centre, University of Surrey, UK, where she is a member of the Nature Inspired Computing and Engineering (NICE) group and the People-Centred AI Institute. She serves as Vice Chair of the IEEE Computational Intelligence Society (CIS) Task Force on Data-Driven Evolutionary Optimization, as a member of the IEEE CIS SHIELD initiative on the ethical and societal dimensions of AI, and as an Associate Editor of Complex & Intelligent Systems. Her research focuses on trustworthy and efficient AI, spanning data-driven and surrogate-assisted evolutionary optimisation, multi-objective Bayesian optimisation, federated and privacy-preserving learning, efficient large language models, and audio deepfake detection. She received her PhD from the University of Surrey under Professor Yaochu Jin, funded by Honda Research Institute Europe, and subsequently held a postdoctoral position at Bielefeld University, Germany. She has published more than 40 peer-reviewed papers in leading venues, including a widely cited survey on Bayesian optimisation in ACM Computing Surveys and work in IEEE Transactions on Evolutionary Computation.
