Safran is an international high-technology group, a leading equipment manufacturer in aerospace, space and defense. With more than 83,000 employees worldwide, Safran develops the technologies that shape the future of air and space mobility.
Safran Tech, inaugurated in 2015 in Magny-les-Hameaux, represents the heart of the Group's innovation. This research center of 600 researchers and engineers concentrates cutting-edge expertise in AI, advanced materials, numerical simulations and disruptive technologies.

Part of the Research and Technology center of Safran Group (Safran Tech), the CASPer laboratory (Confiance, Autonomy, Simulations and Perceptions) is responsible for algorithmic design of autonomy and artificial vision building blocks for autonomous systems in terrestrial and aeronautical domains.
The CASPer laboratory is developing advanced multi-agent AI architectures integrating the lastest GenAI technologies such as Large Language Models (LLM) and GraphRAG systems.
We are particularly interested in creating multi-agent systems with knowledge graphs to enhance the trustworthiness and security of our systems.
This internship is part of applied research effort aimed at creating reliable and interpretable agentic AI systems.

• Multi-Agent AI Systems Development: Design and implement collaborative agent architectures using LLMs for automated analysis and decision-making
• GraphRAG Systems Enhancement: Optimize retrieval and generation pipelines augmented by knowledge graphs for specialized technical domains
• Novel Knowledge-Agent Interaction Architectures: Experiment and test new architectures for knowledge-agent interactions using emerging protocols and quantify the quality of trust and confidence
• LLM Fine-tuning: Adapt foundation models to specific needs of aerospace and critical systems
• Backend/Frontend Development: Create intuitive user interfaces and robust APIs for AI system integration
• User Needs Analysis: Collaborate with engineering teams to identify priority use cases and define technical specifications
• Evaluation and Validation: Develop performance metrics and test protocols to ensure reliability traceability of developed systems
Expected Deliverables
• Functional prototype of a multi-agent AI system with demonstration on real use case
• Complete technical documentation of developed architectures
• Comparative evaluation report of implemented GraphRAG approaches
• User interface enabling interaction with AI agents
• Results presentation and recommendations for future developments