Extract
Digital text, tables, headings, image references, captions, and page boundaries are extracted through multiple engines.
Verified document conversion
PDF to Markdown with page-level proof
Convert complex PDFs into clean Markdown for AI, search, documentation, and archives. Every page gets a quality verdict.
Tables
Figures
Captions
Critical values
Answer plus verdict
Multiple converters. Independent evidence. An honest result.
A normal converter returns text. KYAMOS runs independent extraction paths, compares critical values page by page, and preserves uncertainty instead of hiding it.
Verify
Optional OCR provides independent evidence for scans and difficult pages. It is checked, not accepted blindly.
Decide
Each page receives a score and a traceable verdict: confirmed, likely, uncertain, or failed.
Quality report
You can see what passed and what needs review.
The document verdict is derived from page-level evidence. A plausible-looking page is not enough.
Confirmed
All required checks passed with no unresolved page-level issues.
Likely
Conversion succeeded using a lighter evidence set.
Uncertain
Markdown is available, but one or more pages need review.
Failed
No usable verified result was produced.
Regression proof
Technical annex
- Failed pages
- 0
- Target accuracy
- 99%
- Table page 23
- 1.000
- Mixed page 30
- 1.000
Confirmed regression result recorded 27 July 2026. This is a test proof point, not a claim that every PDF is perfect.
Asynchronous API
Submit, poll, download.
Large documents continue processing after the browser disconnects. The web gateway uses KYAMOS account authentication and user-bound job tokens.
Direct API keys and batch endpoints are available for private pilots.
POST /api/pdfmd/jobs?profile=verified&ocr=auto
Content-Type: multipart/form-data
202 Accepted
{
"job_id": "abc123...",
"status": "queued",
"status_url": "/api/pdfmd/jobs/abc123..."
}
GET /api/pdfmd/jobs/{job_id}
GET /api/pdfmd/jobs/{job_id}/markdown
GET /api/pdfmd/jobs/{job_id}/artifactsManifesto
We do not believe in black-box document conversion.
PDFs contain operational memory: proposals, tables, legal terms, diagrams, and decisions. If a converter silently loses a row or changes a number, it corrupts knowledge. Our promise is clean Markdown, page-level proof, and honest uncertainty.
Operational documents
Built for workflows where silent errors are expensive.
AI retrieval and knowledge bases
Technical proposal archives
Legal and compliance review
Scientific documentation
SaaS document ingestion
Documentation migrations
Private by design
Choose where documents run.
Protected access
Account approval, service tokens, upload limits, rate limits, and per-user job ownership.
Private deployment
Website evaluation jobs expire after seven days. Private deployments can use a customer-defined retention period.
Private beta
Start with an evaluation. Scale to a private deployment.
Pricing is being validated with design partners. Current pilots include conversion, quality reports, and engineering support.
Evaluation
A bounded document and page allowance for product assessment.
Open converterPro pilot
Higher limits, full downloads, quality reports, and API access.
Discuss a pilotEnterprise
Private infrastructure, retention controls, audit support, and service terms.
Discuss a pilotFAQ
Before you upload.
Is this only optical character recognition?
No. Optical character recognition reads text from images. KYAMOS combines multiple PDF conversion engines, optional OCR, and page-level quality checks.
What happens when a page cannot be verified?
The document is marked uncertain. The quality report identifies the affected page, its score, and the unresolved evidence.
Can the service run privately?
Yes. VPN-only, customer-network, and self-hosted deployment are available for private pilots.
Are uploaded PDFs used to train models?
No. Customer files are not used for model training unless a separate written agreement explicitly permits it.