STAGE PFE - chargé d’étude R&D – Explicabilité des réseaux de neurones sur graphes pour le réseau électrique F/H
Core
Develop explainability methods for graph neural networks (GNNs) to interpret defect filtering decisions in a real-time power grid management system.
Role type
Research intern (Machine Learning / Graph Neural Networks)
Builds
Open-source GNN library (EnerGNN) for energy grid applications
Domain
Energy grid operations / Machine Learning / Graph Theory
Deliverable
production ML models
Required skills
Python, Deep Learning, Graph Neural Networks (GNN), JAX, Software Development (Git, Testing), English (fluent)
Preferred skills
Open source contribution, JAX, Explainable AI (XAI), Power grid domain knowledge
Technologies
JAX, Python, EnerGNN, GNNExplainer
Responsibilities
Conduct literature review on GNN explainability; Define explanation targets and metrics for defect filtering; Implement and apply explainability methods (gradients, occlusions, learned masks); Analyze results and propose model improvements; Present findings to domain experts.