36 / VC & PE Funds
Private Intelligence Layer for an Investment Fund
We built an investment fund its own private intelligence layer, trained on the fund's real memos and vocabulary, so meeting prep becomes one cited question instead of a hand-assembled search across 5 systems.
- Client
- NorthPoint Capital: $155M SEBI-Registered Investment Fund
- Stack & approach
- VCPEInvestment OntologyOn-PremiseNorthPoint-OwnedCited Answers
The problem
NorthPoint's hardest material (diligence notes, legal documents, sector research and the reasoning behind a pass) lives as text, in meetings and in people. A schema handles the metrics but not the part that decides investments. Partners meet 5-10 companies daily, and preparation is assembled by hand across several systems every day.
Solution & architecture
A NorthPoint-owned deployment: Omni runs on infrastructure NorthPoint owns, with the knowledge store, document store and connector credentials inside the fund's perimeter. We built a NorthPoint investment ontology from the fund's own memos, internal terms and real partner questions. Collection runs on the research seats the fund already pays for, and every answer names its source, stating what it does not cover rather than fabricating.
Why it is built this way
NorthPoint-owned
Runs on the fund's own account and keys
No licence revocation
No server that can switch it off
Refuses over fabricates
States gaps rather than guessing
