Generalist LLMs are not experts in private credit data constructs – they are models trained on general information on the internet. There is of course value in these models but enabling that value requires a technical strategy on creating domain-specific agents with domain specific vocabulary, skills, knowledge and reasoning.
SEMANTICS: NEL ONTOLOGY
The NEL Ontology is a comprehensive knowledge graph of credit constructs and their relationships.
SYNTAX: LANGUAGE METAPATTERNS
N21 Metapatterns encode senior-level human knowledge of finance language and usage patterns, with the ability to dynamically adapt to previously unseen syntax.
NORMALIZED MEMORY: VECTOR + GRAPH STORAGE
The N21 Data Lake offers a hybrid storage model with normalized data fields across your entire portfolio. Your RAG efforts can be immediately powered by harmonized data and semantic vector embeddings tied to the foundation knowledge graph and source language.
REGISTER for access to deeper strategic and technical content, as well as our periodic white papers.
For this example, the portfolio in the chat has been constrained to Jan-July 2024 of the over 9000 transactions in the N21 Public Dataset.
Production usage scenarios retrieve from private portfolio data, with N21 Public data available for integration.