RAG / Vector Search Engineer
Job Description
About this role
Retrieval-Augmented Generation looks easy in a demo and hard in production, where chunking, hybrid search, and re-ranking decide whether the system actually works. As a RAG / Vector Search Engineer for AI training, you will help AI generate retrieval pipelines that handle real corpora, real queries, and real failure modes.
Key Responsibilities
• Generate and evaluate instruction-response pairs covering chunking, embeddings, and retrieval design.
• Review AI-generated code using pgvector, Pinecone, Weaviate, Qdrant, Milvus, and Vespa.
• Provide feedback on hybrid search (BM25 • dense), re-ranking, and query rewriting.
• Validate AI handling of evaluation harnesses (Ragas, TruLens) and retrieval metrics.
• Evaluate ingestion pipelines, document parsing (PDFs, HTML, OCR), and metadata design.
• Identify subtle issues in embedding drift, stale indexes, and adversarial-document risk.
Ideal Qualifications
• 3• years building production RAG or semantic search systems.
• Deep familiarity with at least two vector databases.
• Strong grasp of embedding models, distance metrics, and vector index trade-offs (HNSW, IVF, ScaNN).
• Experience with hybrid retrieval and modern re-rankers (cross-encoders, ColBERT).
• Comfort with Python and at least one orchestration framework (LlamaIndex, LangChain).
• Familiarity with on-premises and cloud deployment of vector stores is a plus.
Project Timeline
• Start Date: Immediate
• Duration: Ongoing
• Commitment: Flexible, 10-25 hours/week
Contract & Payment Terms
• Independent contractor agreement
• Remote work — anywhere in eligible locations
• Weekly payment via Stripe or bank transfer
• Flexible hours
Get AI past the demo phase of RAG — apply now!