Development of an LLM for interpretation of Technical Documentation on aeronautical systems

Main missions: • User needs analysis: Understand the issues encountered by technical support teams when interpreting complex technical documents and precise technical facts. • Development of the LLM: Contribute to the design and training of a language model adapted to the reading and analysis of technical documents • Citations of sources: Integrate an information traceability functionality, allowing the LLM to provide answers by explicitly citing the documentary sources used. • Testing and validation: Participate in the validation of the model in real conditions on common technical support scenarios (bug resolution, interpretation of technical specifications, etc.). • Continuous improvement: Collaborate with the development team to adjust and optimize the model based on user feedback. The trainee will be asked to participate in demonstrations of their model internally but also externally (Salon Bourget, Eurosatory, etc.)

Job Requirements • Good knowledge of natural language processing (NLP) models and neural networks. • Mastery of Python programming languages ​​and AI-related libraries (PyTorch, TensorFlow, Hugging Face, etc.). • Initial experience with AI tools like llama, GPT, mistral or other language model architectures would be a plus. • Linux, Windows operating system. • Gconf: Git, GitLab • Knowledge of Docker and FastAPI containerization Qualities required: Curiosity, autonomy, knowing how to listen, knowing how to work in a team, spirit of synthesis and analysis