R&D Artificial Intelligence Engineer (M/F)

About the company

Dassault Systèmes, the 3DEXPERIENCE company, is a "human progress accelerator." It offers companies and individuals collaborative virtual environments that allow them to imagine more sustainable innovations. By developing a virtual twin of the real world, thanks to the 3DEXPERIENCE platform and its applications, Dassault Systèmes empowers its customers to push the boundaries of innovation, learning, and production. Dassault Systèmes' 20,000 employees work to create value for our 270,000 customers of all sizes, in all industries, in more than 140 countries. For more information, visit our website www.3ds.com/fr

Joining us also means joining a scientific company at the heart of technological innovation, driven by strong growth for over 40 years.

Your tasks

To maintain the innovative character of the solutions in the DELMIA Generative Experiences AI department, the team is currently looking for an R&D Artificial Intelligence Engineer Intern (LLM, Deep Learning). You will join, in a joyful and creative atmosphere, a team of developers and researchers who design the technological building blocks of solutions for Industry 4.0, and you will experiment with the latest advances in AI.

The integration of language models into industrial workflows poses a recurring question: should we rely on large general-purpose models (LLMs), specialize smaller models (SLMs) through fine-tuning, or augment an existing model via RAG? Each option offers a different compromise between cost, performance, deployability, and data control.

For 6 months, you will empirically and rigorously evaluate these approaches on tasks specific to DELMIA's manufacturing domain (process planning, work instruction generation, exploitation of BOMs and business knowledge). A central question: can specialized SLMs, which are lighter and deployable on-premise, compete with large general-purpose models on targeted tasks?

You will be fully integrated into the R&D team and lead the entire experimental process, under supervised autonomy. Concretely:

• Data curation: you will compile, clean, and structure a representative corpus of industrial data (process documents, BOMs, work instructions, CAD/PLM data), and define the training and evaluation formats adapted to each task.
• RAG pipeline construction: you will implement a solid RAG baseline (indexing, retrieval, possibly reranking), which serves as a reference against fine-tuning approaches.
• Fine-tuning experimentation: you will conduct fine-tuning campaigns for LLMs and SLMs using parameter-efficient techniques (LoRA / QLoRA), exploring hyperparameters, model size, and the volume of data required to reach a target performance level.
• Comparative benchmark: you will design and execute a rigorous head-to-head evaluation of the different approaches (general-purpose LLM, fine-tuned SLM, RAG, hybrid combinations), using metrics covering quality, inference cost, latency, and deployability.
• Synthesis and recommendations: you will formalize a documented decision framework, allowing for the choice of the right approach based on the task type, deployment constraints, and data confidentiality issues.

Requirements

You are in your final year of engineering school or a Master's degree, passionate about computer science and programming, creative, and autonomous.

• Advanced knowledge in Deep Learning, particularly on Transformers and multimodal models.
• Good understanding of LLM-based agent systems.
• Python, PyTorch (or TensorFlow).

Preferred skills

• Mastery of the Hugging Face ecosystem, efficient fine-tuning (PEFT / LoRA), and RAG.
• Knowledge of the manufacturing / PLM domain.
• Git, Docker, Windows, Visual Studio.

We are looking for a dynamic personality, a proactive individual, willing to engage in technical discussions with the rest of the team. Curiosity and perseverance are crucial elements for the success of an ambitious R&D project. Good proficiency in English, both written and spoken, is important.

What we offer

Main advantages and benefits

• Collaborative and innovative environment
• International collaboration
• Diversity of technologies, products, and solutions
• Commitment to diversity and inclusion