Design a Multi-Tenant Enterprise RAG Assistant: Hybrid Retrieval, Reranking, ACLs
Company: OpenAI
Role: Machine Learning Engineer
Category: ML System Design
Difficulty: hard
Interview Round: Onsite
Overview: Design a multi-tenant retrieval-augmented generation assistant that answers employees' questions from their company's documents with citations. Cover ingestion and embeddings, hybrid keyword and vector retrieval with score fusion, reranking, permission-aware multi-tenancy, evaluation, and how to train or adapt embedding models.