Armada AI

AI Assistant retrieval improvements

Contributions to an inherited RAG application, distinct from the Support Agent I later owned.

Contributor. Redesigned the LangGraph workflow and improved context preparation, retrieval and metadata handling.

Problem

The inherited pipeline had grounding failures, duplicate retrieval results, inconsistent metadata, oversized context and chunking defects.

My contribution

Restructured the LangGraph workflow and context preparation, corrected a source-metadata mapping bug, and added query decomposition for multi-intent questions.

Proposed and implemented dense-plus-sparse retrieval, optimized batch ingestion into Qdrant and consolidated separate retrieval indexes.

Delivery boundary

The workflow changes were adopted and deployed. Embedding-service work was completed but not deployed.

Tools

LangGraph · Qdrant · Hybrid retrieval · Query decomposition