Description
You will lead the technical direction of AI components for DocGPT, Ubisoft's internal knowledge assistant platform.
Responsibilities:
- Define and drive the technical direction of retrieval pipelines: chunking, semantic indexing, vector search, reranking.
- Establish evaluation standards for the platform, prioritising response quality.
- Define the integration approach for Large Language Models (LLMs) in a multi-tenant context: context management, prompt engineering, cost control.
- Proactively identify limitations of the current architecture and propose improvements.
- Mentor junior and intermediate team members on best practices in Machine Learning (ML) and experimental rigour.
- Act as a trusted interlocutor for client teams and stakeholders on AI subjects.
Qualifications:
- 5+ years of experience in applied Machine Learning (ML) on information retrieval or Natural Language Processing (NLP) systems in production.
- Mastery of Python and ML/NLP frameworks: LangChain, LlamaIndex, HuggingFace Transformers.
- Proven experience with RAG architectures: embedding models (OpenAI, Cohere, BGE), vector databases (Qdrant, pgvector, Weaviate, Pinecone), reranking (cross-encoders).
- Familiarity with LLM APIs (Claude, OpenAI GPT) and advanced prompt engineering techniques.
- Experience with RAG evaluation frameworks: RAGAS, MRR/NDCG metrics, evaluation dataset construction.
- Knowledge of cloud environments (AWS, Databricks) and containerised environments.
- Demonstrated technical leadership and comfort with ambiguity; excellent communication skills.