RAG Prompt Engineering – Retrieval & Query Design

I will engineer the complete prompt layer for your Retrieval-Augmented Generation (RAG) system — covering query rewriting prompts, retrieval context injection templates, answer synthesis prompts, citation formatting instructions, confidence-based fallback handling, and multi-turn conversation prompts that maintain coherence across follow-up questions. A RAG system is only as good as the prompts orchestrating it; poorly engineered RAG prompts produce answers that ignore retrieved context, hallucinate despite having the right documents available, or fail to handle follow-up questions coherently — undermining the entire purpose of building a knowledge-grounded AI system.

The prompt engineering package covers: query expansion prompts that improve retrieval recall, context window stuffing strategies for maximum relevant content per token budget, synthesis prompts that force grounded responses using only retrieved content, source attribution formatting, graceful handling of insufficient retrieval results, and conversation history compression prompts for multi-turn sessions. All prompts are tested against your knowledge base content and documented with performance observations.

This service is essential for teams building RAG-based chatbots, document Q&A systems, internal knowledge assistants, or customer-facing AI support tools who want the retrieval and synthesis layers of their system to perform reliably at production scale.