Build and Evaluate a Minimal Embedding-Based RAG Pipeline
Company: Harvey
Role: AI Agent
Category: ML System Design
Difficulty: medium
Interview Round: Onsite
Overview: Build a minimal retrieval-augmented generation pipeline with transparent ingestion, chunking, embeddings, indexing, retrieval, and prompt assembly. Evaluate retrieval first with known relevant chunks, preserve document provenance, resist prompt injection from sources, and support abstention.
Read the full Harvey AI Agent interview experience this question came from