Cesar Petrescu

Lab · AI

  • Internal

RADI

An internal research assistant for RADIOCOM procedures and regulations

I developed RADI to help colleagues find answers in company procedures, work instructions and regulations. It combines retrieval-augmented generation with an agentic search process that checks answers against the underlying documents.

  • RAG
  • Hybrid retrieval
  • Reranking
  • AI agents

Document pipeline

  1. Route the question to the relevant company material.
  2. Run multi-hop retrieval and rerank the candidate passages.
  3. Draft an answer from the retrieved evidence.
  4. Review the answer against its sources; return “I don’t know” when the evidence is insufficient.

Outcome

An internal assistant designed to make operational knowledge easier to find while keeping uncertainty explicit.