/artifacts
Engineering memos, reference architectures, and internal tooling.
Agent Evaluation Harness
Internal Tooling • Python / LangSmith
A CI/CD pipeline step that runs synthetic user queries against the RAG agent, measuring precision/recall drift before deployment. Prevents "silent failures" in production.
Hybrid Graph-RAG Architecture
System Design • Neo4j / Vector
A design pattern for grounding LLMs in structured enterprise data. Combines vector search for unstructured similarity with knowledge graphs for structured relationship traversal.