RASS

Retrieval-Augmented Semantic Search platform for enterprise documents

Retrieval-Augmented Semantic Search (RASS)

_Medical Informatics Engineering Summer 2025_

Built a containerized RAG platform for semantic, citation-backed search over long-form enterprise documents. The system emphasizes 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 full RAG pipeline with semantic search capabilities
  • Integrated automated evaluation using RAGAS and TruLens frameworks
  • Built quality gates for grounding accuracy and answer relevance
  • Containerized deployment for production environments
  • Connected system to real user workflows in medical informatics context

Technical Stack

Python, LangChain, Vector Databases, Docker, RAGAS, TruLens, Enterprise Search

Impact

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