← The work

Desktop · Document intelligence

Observation Engine

Turn documents into evidence-backed observations

The Observation Engine is a desktop tool for serious document work. Where most “AI document” products are built for ML engineers, this one is built for the analyst: it reads heterogeneous documents natively, turns them into structured facts, and keeps every value tied to the exact span it came from — so the output is auditable, not just plausible. It feels less like a startup dashboard and more like a refined institutional instrument: a law firm crossed with a Bloomberg terminal. It ships as a portable app you run on your own machine.

The problem

Analysts copy figures out of Word, Excel, PDFs, and slide decks by hand into spreadsheets. It is slow, error-prone, and impossible to audit a number back to the document it came from.

Who it’s for

Economists, auditors, and researchers who work from piles of heterogeneous documents and need every figure traceable to its source.

What it does

  • Reads DOCX, XLSX, PDF, and PPTX natively
  • Turns documents into structured, evidence-linked facts
  • Every value traces back to its exact source span
  • Entity linking and confidence-ranked review queues
  • Runs on your own machine — Windows and Mac

Built with TypeScript · SQL Server · Python

Built in the Vanplant studio.

Want one of these for your business?

Build with us