Skill

RAG (Retrieval-Augmented Generation)

Grounding LLM answers in retrieved documents — I've built RAG pipelines in Python and a provider-neutral RAG package in Dart.

  • #ai
  • #retrieval

Retrieval-augmented generation (RAG) retrieves relevant chunks from a document store — usually by embedding similarity — and passes them to an LLM as context, so answers are grounded and citable instead of relying on model memory.

How I use it

  • local-rag-chatbot — a fully local pipeline: idempotent ingestion of Markdown/text/PDF, recursive chunking, CPU embeddings, a Chroma vector store, multi-query expansion, reciprocal rank fusion, a grading step that falls back to "I don't know", and answers with source citations. Orchestrated with LangGraph.
  • Akashi — the akashi_rag package adds a provider-neutral Retriever seam, an in-memory vector store, document chunking and retrieval-as-tool.
  • Somnio Software — RAG and pgvector are part of the stack in my current role.

Credentials and courses

AI Agents Using RAG and LangChain and Generative AI Applications Specialist in the IBM Generative AI Engineering Professional Certificate certificate; Building Your First AI Agent with LangChain in the Agentic AI Engineering Specialization specialization.

Related