Connected Knowledge System for a Large Research Organisation

Problem

A large research organisation published dozens of policy reports per year — each containing findings, arguments, and recommendations — but they accumulated as PDFs that went unread after launch week. Teams had no way to query institutional knowledge across publications.

Approach

Designed and deployed a Connected Institutional Knowledge System (CKS) grounded in the organisation's full publication corpus. Built a hybrid retrieval pipeline combining graph-based knowledge extraction with Gemini-powered claim extraction, epistemic provenance flags distinguishing observed findings from analytical inferences, and a conversational interface. All processing stays within the organisation's environment.

Result

Researchers query 94 publications in natural language and get sourced answers in seconds with inline citations. Every claim carries its epistemic status — observed or inferred — so teams know what to trust. A D3.js force-directed graph visualises how concepts connect across projects.

• ◆94 publications ingested and searchable from a single interface
• ◆Every claim flagged as observed or inferred, preventing institutional mythology
• ◆Cross-project concept connections visible in an interactive knowledge graph
• ◆Gemini 2.0 Flash extracts structured claims with verbatim source quotes