保険業務の判断精度を高め、損害査定プロセスを自動化します
Built on Brix, the system searches insurance terms, regulations, and case law in real-time to generate evidence-based answers. Operates on on-premise GPU to ensure zero data leakage
Live RAG Chatbot Interface
Blurred to protect client information
PNP Loss Adjustment — On-premise RAG Chatbot
PNP RAG
Built a Brix-powered RAG chatbot for PNP Loss Adjustment. Evidence-based answers generated from real-time search across policy terms, regulations, and case law on air-gapped on-premise GPU infrastructure
93%
Accuracy
Answer accuracy for compensation queries
On-prem
Infrastructure
Air-gapped GPU, zero data leakage
5mo
Duration
Django, Qdrant, On-prem GPU
Document Search
Real-time vector search across insurance terms, regulations, and case law
Evidence-based Answers
Answers with explicit source citations to support adjuster judgment
Air-gapped
Sensitive insurance data never exposed to external networks