Cortexio Lens turns contracts, reports and scanned documents into structured, verifiable data.
You point Cortexio Lens at a messy document — a contract, a financial report, a scanned PDF, a form. It handles the layout, tables, stamps and poor scans, and returns the fields, tables and clauses that matter as structured data. Every value links back to where it appears in the source. A person checks low-confidence values before anything is released.
Built for the teams that live in documents — finance, insurance and operations.
Drop in PDFs, scans, photos and emails — one document or a whole batch.
Lens reads the page the way a person would: layout, tables, stamps, poor scans.
A vision-language model maps what is on the page to what it means.
Fields, tables and clauses come out structured, each with a confidence score.
A person checks low-confidence values; every value links back to the source.
Structured data flows into your systems through an API or export.
Over time, your corrections will tune Lens for that document type, where you work.
Contracts, reports, forms and photographed documents — Lens handles the layout and the mess a flat text extractor trips over.
Early accessEach field, table and clause links back to where it appears in the document, so a reviewer can check a value in one click — no blind trust.
Early accessThe structured result flows into downstream systems through an API or export. Over time, your corrections will tune Lens for the documents you see most.
Early access · API & exportLens is honest about what it is unsure of. High-confidence values pass; the doubtful ones are flagged for a human.
Every value links back to where it appears in the source, so a check takes seconds, not a re-read.
Set which fields always need a human sign-off, however confident the model is.
Today, Lens reads and extracts through cloud model APIs, and a person checks low-confidence fields before anything is released. The step that makes it fast and private is bringing the reading model onto NVIDIA GPUs of our own.
Today Lens runs on cloud model APIs for reading and extraction, with a person checking low-confidence values before release. No GPUs of our own yet.
Next, we will fine-tune our own vision-language model on document layouts on NVIDIA GPUs with CUDA and PyTorch. We will serve it with NVIDIA Triton Inference Server and NVIDIA NIM, and use NVIDIA TensorRT so a long contract is parsed in seconds.
Today: cloud model APIs. Next: our own model on NVIDIA GPUs.
Cortexio Labs is an independent software company building Cortexio Lens. Privately held, early access, and based in Australia.
| Company | Cortexio Labs |
| Product | Cortexio Lens — document understanding engine |
| Stage | Early access |
| Location | Australia |
| Status | Privately held · no outside funding |
| Contact | shuman@cortexiolabs.com |
Builds Cortexio Lens: the reading models, the extraction engine, and the review loop.
Cortexio Lens is built by Cortexio Labs, Australia.
Request early access and we will invite you to run your own documents through Cortexio Lens.