Onulipi · Document intelligence platform
Every document read, verified and in your systems. No manual keying.
Onulipi identifies each incoming document, extracts the fields it needs, checks them against your rules and against each other, and sends only the uncertain ones to a person. Clean, verified data flows into your loan origination, claims, CRM, ERP, EHR and document systems within minutes, and Onulipi learns from every correction.
Onulipi (Bangla for “copy”): every field, a faithful copy of its source.
- No unverified field reaches your systems
- Every value linked to its place on the page
- Your documents never train shared AI models
- Processing and storage in the region you choose
Grounded values
Linked to source
Each value links to the exact page and region it came from
Your best people are still reading documents and retyping them.
Loan files, claims, contracts and clinical records are still keyed by hand.
Month-end
Only overtime clears it
Capacity tied to headcount
Every peak means overtime.
When a typo surfaces
Costly to fix by then
Errors found too late
A typo surfaces weeks later.
Template OCR
- Form, usual layoutRead
- Form, new layoutFailed
- Contract clausesFailed
- Clinical notesFailed
Back to keying by hand
Old automation breaks
Templates fail on new layouts.
Master agreement
Change of control clause
Somewhere in the prose…
Hours to find one clause
Experts stuck searching
The answer is buried in prose.
Where Onulipi fits
Onulipi removes manual keying for the documents you choose, and puts your people only where their judgement is needed.
Uncertain fields go to a person. Everything else flows straight through.
A field is delivered only when it is confident, passes every rule and links to its source.
The delivered-data guarantee
Our contracts commit to the accuracy of the data we deliver, not to raw model accuracy.
Capture
Read
Classify
Extract
No: a person checks it
Only that one field
Yes: delivered
Into your core systems
Learning loop
The next model, re-tested before release
Inside Onulipi
Every step a document takes.
Capture and read
Any channel, any format, even the bad scans.
Mixed packets are split, related files grouped into one case, and every page cleaned up before it is read.
- Every inbound channel
Dedicated email addresses for each workflow, web upload with drag-and-drop and bulk upload, a REST API, watched folders and cloud storage (SFTP, S3, Azure Blob, SharePoint, Google Drive), and a mobile capture SDK.
- Every common format
Native and scanned PDF, multi-page TIFF, JPEG, PNG, HEIC, Word, and email files (EML and MSG), including attachments and the message body.
- Packets and cases
A 60-page file is split into its documents, and several files are grouped into one case.
- Better images before reading
Deskew, de-noise, de-warp, and contrast and orientation correction. The mobile SDK checks edges, blur and glare before a photo is submitted.
- Reading that understands layout
OCR with reading order, tables and key-value pairs; checkbox states; signature detection; and handwriting recognition for block letters and cursive.
- A quality gate
Each image gets a quality score. Below the threshold, Onulipi asks for a new capture or routes the document to review.
- Mixed languages
The language is detected for each page and region, including pages that mix languages.
- No double processing
Exact and near-duplicate documents are detected before they cost you twice.
Value on day one, in the documents your industry runs on.
Each industry ships as a pack on the same core, so you start from a working configuration.
Lending
- Reads
- Applications, pay slips, statements, IDs
- Checks
- Income and identity across documents
- Delivers
- A case ready for underwriting
Insurance claims
- Reads
- Claim forms, estimates, medical bills
- Checks
- Policy active, amount within limits
- Delivers
- A claim triaged for handling
Legal and regulatory
- Reads
- NDAs, master agreements, leases
- Checks
- Parties, dates, the amendment chain
- Delivers
- A clause register and risk dashboard
Health records
- Reads
- Lab results, discharge summaries, referrals
- Checks
- Patient match, ICD-10 and LOINC codes
- Delivers
- Records in your EHR through HL7 or FHIR
From the applicant's email to the underwriter's desk, the same day.
One 40-page loan file, and the single field a person had to check.
Lending inbox
40 pages
Application form and ID
Three pay slips, two tax returns
Six months of bank statements
Emailed to the lending team's Onulipi address.
Built for operations, and for financial, health and legal data.
Each leader gets something different, on your most sensitive data.
CEO
More volume, same team
COO
Lower cost per document
CFO
Cost per document, monthly
Risk & compliance
Only verified data delivered
CIO/CTO
API-first, region you choose
Your experts
Clauses and figures found
99.9%
monthly availability
10×
daily load headroom
0
received files lost
AES-256
encryption at rest
Also built in
- Every value linked to its source
- Zero-retention AI providers
- No training without your opt-in
- Single-tenant private cloud
- Personal data redacted
- SSO and SCIM
- Full audit log
- SOC 2 Type II controls, audit under way
- HIPAA-ready, with BAAs
- GDPR-ready
- WCAG 2.1 AA
Pay for the pages you process.
A platform fee plus a per-page rate, on the plan that fits your volume.
Starter
Small teams and pilots
- Standard document types
- Prepaid page bundles
Business
Mid-market operations teams
- Custom types, connectors, SSO
- Tiered per-page rate
Enterprise
Banks, insurers, health, law
- Private cloud, SLAs, custom packs
- Committed annual volume
Add-ons
- Industry packs
- Extra environments
- Managed review by our reviewers
Prove it on one flow. Then expand.
Savings measured against your own baseline, not an industry average.
- 1
Pick one flow
Measure its cost and error rate today.
- 2
Configure and test
Your rules, tested on your samples.
- 3
Pilot in production
Live documents, agreed targets.
- 4
Expand
New types by configuration.
Questions operations leaders ask us.
We do not promise that models never err; no honest vendor can. We promise that no unverified field reaches your systems. A field is delivered automatically only when its calibrated confidence clears your threshold, every validation rule passes and it links to its source. Everything else is checked by a person. Blind sampling measures the real error rate in production, and our contracts commit to the accuracy of the data we deliver.
Get started with Onulipi
See Onulipi read your documents, not ours.
Choose one document flow that costs you the most to process by hand. Send us a sample set under NDA, and we will show your team the straight-through rate, the fields that need review and the verified output, on your own documents.
Prefer to pick a time yourself? Schedule a meeting