RASS

Retrieval-Augmented Semantic Search (RASS)

Built a containerized, multi-service RAG platform for semantic, citation-backed search over long-form enterprise documents. The system decouples asynchronous ingestion from query serving and treats evaluation as a first-class requirement, with automated quality gates for grounding and relevance integrated into the deployment pipeline.

Key Contributions

• Designed and implemented a multi-service RAG pipeline using Node.js, Docker, OpenSearch, Redis, and PostgreSQL
• Combined hybrid search and reranking for citation-backed retrieval over more than 3,000 enterprise documents
• Measured more than 85% relevance with RAGAS and TruLens pre-deployment quality gates
• Exposed REST, MCP, and SSE interfaces with structured citations, JWT/API-key authentication, user-scoped security filters, and OpenTelemetry observability
• Decoupled asynchronous document ingestion from query serving for up to 25 concurrent users and AI agents

Technical Stack

Node.js, Docker, OpenSearch, Redis, PostgreSQL, RAGAS, TruLens, Enterprise Search

Impact

• Enabled citation-backed semantic search across thousands of pages of technical documentation
• Established an evaluation-first approach to ensure answer quality before deployment
• Provided measurable quality and observability signals for stakeholder confidence

Links

• RASS / CoRAG repository