PhD Position in Cognitive Network Observability: Integrating GNNs and LLMs

This PhD position at Inria’s Rennes center focuses on advancing cognitive network observability by merging Graph Neural Networks (GNNs) with Large Language Models (LLMs). Hosted by the ERMINE project-team in collaboration with ADOPNET, the research aims to bridge mathematical inference with contextual reasoning to enable explainable, generalizable network tomography. The PhD student will design novel GNN architectures for zero-shot topology transfer, integrate Retrieval-Augmented Generation (RAG) frameworks for explainable diagnostics, and validate the framework on real-world network data. The project is part of the GENIE ANR consortium, which seeks to revolutionize network management through AI-driven automation.

• Design and implement advanced Relational Graph Convolutional Networks (RGCNs) for inferring non-additive network metrics and achieving zero-shot transfer across unseen topologies.
• Develop a Retrieval-Augmented Generation (RAG) framework tailored for network graphs, incorporating historical incident data, routing logs, and vendor documentation into a vector database.
• Integrate GNN inference pipelines with LLMs to generate human-readable, context-aware root cause diagnostics for network anomalies.
• Conduct extensive experiments on real or highly realistic network data to validate the generalizability and robustness of the proposed framework.
• Publish research findings in international peer-reviewed conferences and journals, and present results at scientific events.
• Collaborate with partners in the GENIE ANR consortium and participate in project meetings and research activities.