Apprentissage ingénieur AI Developer ( H/F)
Description & Requirements
Job Description
About the Project
We are developing SAM (Smart Assistant for Maintenance), an AI-powered assistant designed to help maintenance teams quickly access Standard Operating Procedures.
SAM will leverage modern AI technologies such as Retrieval-Augmented Generation (RAG), local Large Language Models, and semantic search to transform how technical documentation is accessed and used in daily operations.
The role:
We are looking for an AI Engineer / Developer to design, build, and deploy the SAM solution. You will work on integrating local AI models, building document retrieval pipelines, and creating a user-friendly interface that enables maintenance teams to interact with SOP documentation efficiently.
Key Responsibilities
- Design and implement an AI assistant using RAG architecture
- Integrate local LLMs (e.g., via Ollama or equivalent solutions)
- Build document ingestion and semantic search pipelines
- Implement a secure internal knowledge database for SOP documents
- Ensure answers are traceable to source documents
- Optimize system performance, reliability, and security
Required Skills
- Experience working with AI/LLM solutions and open models (e.g., models released by Meta)
- Knowledge of Retrieval-Augmented Generation, embeddings, and vector databases (e.g., Chroma or similar)
- Familiarity with LLM orchestration tools such as LangChain
- Experience processing technical documents (PDFs, manuals, procedures)
- Ability to build lightweight interfaces (e.g., web, mobile or reactive)
- Understanding of secure or on-premise deployments
Nice to Have
- Experience in manufacturing, maintenance, or industrial environments
- Experience deploying AI solutions in secure corporate infrastructure
- Knowledge of knowledge-management or documentation systems
What Success Looks Like
- Maintenance teams can retrieve SOP information through natural language queries
- Answers are reliable, grounded in internal documentation, and traceable
- SOP search time is significantly reduced
- The system is scalable to thousands of documents