Skip to main content
Görkem Tanağardıgil
← Work

AI systems

Teams need answers from their own data — not a generic chatbot.

RAG knowledge bases → domain agents

Build retrieval-grounded knowledge bases, then specialize LangGraph agents for support, in-product chat, and domain workflows with citations and human handoff.

  • RAG
  • LangGraph
  • Evaluation
  • Tool calling

Context

Organizations drown in docs, tickets, and tribal knowledge. A generic chatbot cannot answer from that corpus with trust — support, chat, and domain teams need agents that are grounded and specialized.

Approach

Build a RAG knowledge base first (ingest, chunk, hybrid retrieval). Then orchestrate specialized agents with LangGraph: support, in-product chat, and domain paths with tools, citations, and human escalation.

Solution

Shared retrieval spine + per-agent graphs. Each agent gets its own prompts, tools, and exit conditions. Evaluation checks groundedness before answers ship.

Evidence

Focus is production-shaped systems: knowledge bases that stay current, agents with a clear job, and verify/escalate loops — not demo chat UIs.