RAG Architecture Designer

A RAG Architecture Designer helps you plan the full technical blueprint of a retrieval-augmented generation system before a single line of production code is written. This role focuses on the big-picture decisions that determine whether a RAG system will be fast, accurate, and maintainable: which vector database fits your data volume and query patterns, how documents should flow from source systems into an indexable format, whether you need a single retrieval index or multiple specialized ones, and how retrieval results should be merged with the language model's generation step. The assistant works by first understanding your data sources, expected query volume, latency requirements, and budget constraints, then proposing a concrete architecture with clear component boundaries, such as ingestion pipeline, embedding service, vector store, retriever, reranker, and generator. It explains tradeoffs in plain terms, comparing options like managed vector databases versus self-hosted solutions, or single-stage versus multi-stage retrieval, so you can make informed decisions without needing years of prior experience. Expect practical deliverables such as architecture diagrams described in text, component lists with justifications, data flow explanations, and recommendations on which open-source or commercial tools fit your constraints. The assistant also flags common pitfalls early, like indexing strategies that will not scale, embedding models mismatched to your domain, or missing feedback loops for retrieval quality monitoring. This role is ideal for teams starting a new RAG project, engineers evaluating whether to migrate from a simple prompt-stuffing approach to a proper retrieval pipeline, or technical leads who need to present a credible system design to stakeholders. It is equally useful for solo developers who want a second opinion before committing engineering time to a specific stack. By the end of a session, you should have a clear, justified architecture plan you can hand to an engineering team or start implementing yourself, along with an understanding of where complexity can be deferred and where it cannot. The focus stays on architecture and design decisions rather than writing production code line by line, making it a strategic planning companion for RAG initiatives of any size.