3 days ago
Toulouse, FranceSenior
Responsibilities
- Post-train open language and multimodal models through supervised fine-tuning, preference alignment, reinforcement learning, and distillation.
- Build reproducible model-training and retraining pipelines for Airbus engineering tasks.
- Design and maintain evaluation suites covering accuracy, reasoning, hallucination, robustness, and safety.
- Optimize models for on-premise deployment through quantization, pruning, and inference tuning.
- Release tested and versioned models with documentation, model cards, known limitations, and usage contracts.
- Collaborate with agent and engineering teams to define model requirements and measure release impact.
Requirements
- Master’s degree in computer science, machine learning, or a related field.
- Several years of hands-on experience fine-tuning or training deep learning models and shipping them to real users.
- Strong Python and PyTorch skills with experience in the Hugging Face ecosystem, including tools such as Transformers, TRL, and PEFT.
- Experience building ML model evaluations using benchmarks, test sets, and relevant metrics.
- Experience with version control, testing, experiment tracking, and reproducibility.
- Fluent English.
- Preferred experience includes LLM preference alignment or reinforcement learning, distillation, quantization, inference engines, LLM evaluation frameworks, distributed training, and engineering, manufacturing, or aerospace data.
- German language skills are preferred.
Benefits
- Permanent employment with Airbus.
- Work based in Munich, Madrid, Toulouse, Paris, or Hamburg.
- Opportunity to shape AI use in aircraft engineering with visible impact on real products.
- Work with open models, sovereign on-premise infrastructure, and real engineering data.
- Professional development opportunities within the Airbus group.
- Flexible working arrangements are supported where possible.
