CaveauAI

CaveauAI is Blue Note Logic's flagship document intelligence platform. It transforms unstructured document collections into queryable knowledge bases using large language models running entirely on owned infrastructure.

Unlike cloud AI services where your data traverses third-party servers, every CaveauAI query runs on GPUs we own in EU data centres we chose. The platform indexes documents into 768-dimension semantic embeddings, retrieves the most relevant passages, and generates answers with precise citations back to source material.

How CorpusAI Works

The platform follows a three-stage pipeline: ingest, embed, query. Documents are parsed and chunked into semantically meaningful segments. Each chunk is embedded into a 768-dimension vector space using models optimised for retrieval. When a user asks a question, the system retrieves the most relevant chunks, feeds them to a 72B parameter reasoning model, and returns an answer with page-level citations.

Architecture

• Embedding Pipeline: 768-dimension semantic vectors stored in PostgreSQL with pgvector
• Inference: 72B parameter models running on NVIDIA RTX PRO 6000 Blackwell GPUs
• Networking: WireGuard mesh connecting all nodes with encrypted point-to-point tunnels
• Isolation: Cryptographic tenant separation — each corpus is an encryption boundary
• Storage: Documents never leave the EU; all processing happens on owned hardware

Use Cases

CorpusAI is deployed across legal research, regulatory compliance, healthcare documentation, financial analysis, and construction project management. Any organisation with thousands of documents and questions that need answers — not summaries, not chat, but verified, cited answers — is a CorpusAI customer.