Home / Articles
Series · 5 parts
Knowledge graphs in private credit
Why private credit exposure lives in the links between borrowers, sponsors, lenders and holders, and how to model it. Part 2 builds a real graph from seven reconciled BDC 10-Qs.
- PART 1Why private credit is a graph problemPrivate credit exposure lives in the links between borrowers, sponsors, lenders, facilities, guarantors and holders, not in any single row. Here is why spreadsheets break and what a graph model fixes.
- PART 2We built a cross-lender borrower graph from 5,867 BDC holdingsUsing SEC schedule-of-investments data for seven reconciled BDCs at 6/30/2026, we built a small networkx graph of BDCs and borrowers. After filtering, 208 of 2,034 approximately matched borrowers are held by two or more of the seven.
- PART 3Knowledge graphs + LLM agents (GraphRAG) for credit agreementsCredit agreements are networks of defined terms, cross-references and baskets. Representing them as a graph gives LLM agents a better way to answer questions like 'can this borrower move assets to an unrestricted sub?' than chunk-based retrieval alone.
- PART 4Graph use cases by sector: CLOs, secondaries, sponsors, fund finance and SRTFive concrete graph patterns for credit teams: CLO look-through overlap, secondaries portfolio overlap, sponsor-network contagion, LP concentration in fund finance, and SRT reference pool analysis. Each covers what the nodes and edges are, the query, and where the data comes from.
- PART 5Build vs buy: graph tools for credit teamsA neutral, sourced overview of graph databases (Neo4j, Kuzu and its forks, Memgraph, TigerGraph, Amazon Neptune) and entity-resolution data sources for private credit teams, with the questions to ask before you build or buy. No rankings.