AI business process automation with MCP and a RAG knowledge base
MY ROLEProject lead, AI/software architect and backend developer
As project lead, AI/software architect and backend developer, I implemented an AI-assisted business process with controlled tool calls and migrated a document-based knowledge base to a new operating environment.
Backend APIs and MCP tools connected the agent to existing systems. JSON Schema-based data contracts and explicit human approvals controlled tool calls. Repeatable evaluations and validation without operational write access prepared the introduction.
A document-based knowledge base was migrated to a new environment using Podman and Quadlet. PostgreSQL and pgvector provided the foundation for the document index and HNSW-based vector search. Indexing, retrieval and reranking were compared before and after migration using the same test cases. Backups, restoration and rollback to the previous environment were tested.
Result: Implemented an MCP-supported business process with approvals and evaluation, and migrated a RAG knowledge base with verified search and restoration.