NGT Unstructured Data Extraction
Ingest fund admin reports, custodian statements, and other structured data buried in unstructured documents.
NGT Unstructured Data Extraction ingests documents that carry operational data in formats that resist direct processing: PDF fund administrator reports, Excel outputs, custodian statements, and other files where the data exists but is not machine-readable as delivered.
What it does
The tool parses incoming documents, identifies the fields relevant to each document type, extracts structured values, and validates them against expected schemas. Outputs route to the SPARK data layer, where they become inputs for NGT Tools analytics, reconciliation controls, and investigation workflows.
The extraction process is deterministic. Each document type has a defined extraction spec. Results are auditable: the source document, the extracted values, and any validation flags are all retained.
Supported document types
- Fund administrator NAV reports and accounting statements
- Custodian position and cash statements
- Transfer agent flow reports
- Fee and expense schedules
- Custom document types configurable per client
Agent access
The tool is callable by agents through MCP and APIs. An agent processing an investigation can pull extracted data directly from source documents without manual intervention.