# Quarry

> Quarry sources, licenses and verifies proprietary data for frontier AI labs, rights-cleared and ready to train on.

Quarry is a broker and exchange, not a data owner. It matches lab demand to owners (BPOs, call centres, hospitals, diagnostic chains, platforms), runs de-identification at the source so no personal data leaves the owner's systems, and handles licensing, QA and delivery.

## Site

### Data
- [Data overview](https://quarry-website.vercel.app/data): What Quarry supplies, and how labs get it
- [Conversational audio](https://quarry-website.vercel.app/data/conversational-audio): Real calls in Indian and global languages
- [Medical imaging](https://quarry-website.vercel.app/data/medical-imaging): CT, MRI, mammography and X-ray with reports
- [Catalogue](https://quarry-website.vercel.app/catalog): Every licensable dataset, as a deal sheet
- [Demand board](https://quarry-website.vercel.app/demand): What frontier labs are asking for

### Data owners
- [Sell your data](https://quarry-website.vercel.app/data-owners): License your archive without data leaving
- [Value your data](https://quarry-website.vercel.app/data-owners/value): A two-minute, private estimate
- [Demand board](https://quarry-website.vercel.app/demand): Open requests you could fill
- [Refer an owner](https://quarry-website.vercel.app/refer): Get paid for the introduction

### Trust
- [Trust center](https://quarry-website.vercel.app/trust): Can we sign with Quarry? One screen
- [De-identification and privacy](https://quarry-website.vercel.app/trust/deidentification): What leaves, and how risk is tested
- [Security](https://quarry-website.vercel.app/trust/security): Infrastructure, access, incident response
- [Provenance and licensing](https://quarry-website.vercel.app/trust/provenance): Rights review and chain of title
- [Compliance](https://quarry-website.vercel.app/trust/compliance): Status per framework, honestly labelled
- [Subprocessors](https://quarry-website.vercel.app/trust/subprocessors): Who processes data for Quarry
- [Request documents](https://quarry-website.vercel.app/trust/documents): Questionnaires, policies, DPA

### Docs
- [Docs home](https://quarry-website.vercel.app/docs): The reference both sides read
- [Get started](https://quarry-website.vercel.app/docs/start): How licensing and listing work
- [Data specs](https://quarry-website.vercel.app/docs/specs): Formats, metadata, quality scoring
- [De-identification](https://quarry-website.vercel.app/docs/deid): At-source architecture and reports
- [Provenance and licensing](https://quarry-website.vercel.app/docs/provenance): Manifest schema, licence types
- [Delivery](https://quarry-website.vercel.app/docs/delivery): Methods, packaging, checksums
- [For data owners](https://quarry-website.vercel.app/docs/owners): Onboarding, the at-source agent, payouts
- [API and agents](https://quarry-website.vercel.app/docs/api): Catalogue, demand, MCP, llms.txt
- [Changelog](https://quarry-website.vercel.app/docs/changelog): What changed, newest first

### Company
- [About Quarry](https://quarry-website.vercel.app/about): Why we exist, and how it works
- [Research](https://quarry-website.vercel.app/research): Notes from the supply side
- [Careers](https://quarry-website.vercel.app/careers): Work with us
- [Press](https://quarry-website.vercel.app/press): Wordmark, boilerplate, contact
- [Contact](https://quarry-website.vercel.app/contact): Talk to a person

### Legal
- [All policies](https://quarry-website.vercel.app/legal): Every policy and agreement, versioned
- [Terms of Service](https://quarry-website.vercel.app/legal/terms)
- [Privacy Policy](https://quarry-website.vercel.app/legal/privacy)
- [Data Licensing Terms](https://quarry-website.vercel.app/legal/data-licensing)
- [Sample NDA](https://quarry-website.vercel.app/legal/sample-nda)
- [Supplier Terms](https://quarry-website.vercel.app/legal/supplier-terms)
- [Data Processing Addendum](https://quarry-website.vercel.app/legal/dpa)
- [Acceptable Use](https://quarry-website.vercel.app/legal/acceptable-use)
- [Cookies](https://quarry-website.vercel.app/legal/cookies)
- [Referral Terms](https://quarry-website.vercel.app/legal/referral-terms)
- [Takedown and deletion](https://quarry-website.vercel.app/legal/takedown)
- [Grievance officer](https://quarry-website.vercel.app/legal/grievance)

## Docs

### Get started
- [What is Quarry](https://quarry-website.vercel.app/docs/start/what-is-quarry): A broker and exchange that licenses hard-to-get data from the businesses that hold it to frontier labs.
- [How licensing works](https://quarry-website.vercel.app/docs/start/how-licensing-works): For labs: from a request or a catalogue listing to a signed licence and a delivered dataset.
- [How listing works](https://quarry-website.vercel.app/docs/start/how-listing-works): For owners: from an application to a listed deal sheet, a licence and a payout.
- [Glossary](https://quarry-website.vercel.app/docs/start/glossary): The terms used across Quarry's docs, deal sheets and agreements.

### Data specs
- [Conversational audio](https://quarry-website.vercel.app/docs/specs/conversational-audio): Audio formats, channel layout, the transcript JSON format and metadata fields for call recordings.
- [Medical imaging](https://quarry-website.vercel.app/docs/specs/medical-imaging): DICOM packaging, retained and removed tags, report format, pseudonymous IDs and the study manifest.
- [Metadata dictionary](https://quarry-website.vercel.app/docs/specs/metadata-dictionary): Every metadata field Quarry delivers, with its type and allowed values.
- [Quality scoring](https://quarry-website.vercel.app/docs/specs/quality-scoring): The checks every record passes before delivery, and how results are reported.

### De-identification
- [At-source architecture](https://quarry-website.vercel.app/docs/deid/at-source-architecture): Where de-identification runs, what leaves the owner's environment, and who can access what.
- [Audio and transcript redaction](https://quarry-website.vercel.app/docs/deid/audio-transcript-redaction): How personal data is found and removed from call audio and transcripts.
- [DICOM and report de-identification](https://quarry-website.vercel.app/docs/deid/dicom-report-deidentification): Header cleaning to PS3.15, burned-in text, report redaction and consistent date shifts.
- [Re-identification risk testing](https://quarry-website.vercel.app/docs/deid/reidentification-risk-testing): How residual risk is assessed per dataset before anything is released.
- [Reports you receive](https://quarry-website.vercel.app/docs/deid/reports-you-receive): What a de-identification report contains, and when you get it.

### Provenance and licensing
- [Provenance manifest schema](https://quarry-website.vercel.app/docs/provenance/manifest-schema): One JSON record per delivered item: where it came from, under which licence, and which checks it passed.
- [Licence types](https://quarry-website.vercel.app/docs/provenance/licence-types): Exclusive and non-exclusive licences, term, territory and the sample NDA.
- [Permitted-use matrix](https://quarry-website.vercel.app/docs/provenance/permitted-use-matrix): What a licence lets you do with the data, and what it never does.
- [Datasheet template](https://quarry-website.vercel.app/docs/provenance/datasheet-template): The datasheet every dataset ships with, section by section.

### Delivery
- [Delivery methods](https://quarry-website.vercel.app/docs/delivery/methods): Per-buyer buckets or signed, expiring URLs. Never email attachments.
- [Packaging and formats](https://quarry-website.vercel.app/docs/delivery/packaging-and-formats): How a delivery is laid out on disk, for audio and for imaging.
- [Checksums and verification](https://quarry-website.vercel.app/docs/delivery/checksums-and-verification): How to check that every file arrived whole and matches its manifest record.
- [Deletion and revocation requests](https://quarry-website.vercel.app/docs/delivery/deletion-and-revocation): What happens when an owner or an individual asks for data to be withdrawn.

### For data owners
- [Onboarding checklist](https://quarry-website.vercel.app/docs/owners/onboarding-checklist): What to have ready, from rights paperwork to a sample export.
- [Installing the at-source agent](https://quarry-website.vercel.app/docs/owners/at-source-agent): Run Quarry's de-identification inside your environment, as a container or an on-prem agent.
- [Export formats we accept](https://quarry-website.vercel.app/docs/owners/export-formats): What to export from your call platform or PACS, and how.
- [Payouts](https://quarry-website.vercel.app/docs/owners/payouts): How the revenue share on each licence reaches you.

### API and agents
- [Overview](https://quarry-website.vercel.app/docs/api/overview): Public, read-only endpoints for the catalogue and demand board, plus feeds for agents.
- [Authentication](https://quarry-website.vercel.app/docs/api/authentication): Which endpoints are public, which need an API key, and how keys are issued.
- [Catalogue API](https://quarry-website.vercel.app/docs/api/catalog): List and fetch licensable datasets: GET /api/v1/catalog and /api/v1/catalog/{id}.
- [Demand API](https://quarry-website.vercel.app/docs/api/demand): Read open lab requests: GET /api/v1/demand.
- [Requests API](https://quarry-website.vercel.app/docs/api/requests): Submit a data request and ask for samples from code.
- [MCP server](https://quarry-website.vercel.app/docs/api/mcp-server): Let an agent search the catalogue and demand board over the Model Context Protocol.
- [llms.txt](https://quarry-website.vercel.app/docs/api/llms-txt): Plain-text guides to Quarry for language models: /llms.txt and /llms-full.txt.

### Changelog

## Catalogue (sample listings, illustrative)

Machine-readable: https://quarry-website.vercel.app/catalog.json and https://quarry-website.vercel.app/api/v1/catalog

- QD-7K2P · Health insurance sales calls, five Indian languages · Conversational audio · Sample ready
- QD-3MX8 · Credit card collections and recovery calls · Conversational audio · Listed
- QD-9TQ4 · Food-delivery order-issue calls with IVR handoff · Conversational audio · Under NDA
- QD-4HV1 · Patient appointment and reminder calls · Conversational audio · Listed
- QD-8CR6 · Contrast CT studies with radiology reports · Medical imaging · Sample ready
- QD-2BN5 · Brain and spine MRI, all sequences · Medical imaging · Listed
- QD-6WF3 · Screening mammograms with BI-RADS reports · Medical imaging · Licensed
- QD-5JL9 · Chest X-rays with radiology reports · Medical imaging · Sample ready

## Demand board (example requests, illustrative)

Machine-readable: https://quarry-website.vercel.app/demand.json and https://quarry-website.vercel.app/api/v1/demand

- RQ-0142 · Health insurance sales calls in Indian languages · Conversational audio · 10,000+ hours · needed by Q1 2027 · Open
- RQ-0143 · Patient appointment calls · Conversational audio · 2,000+ hours · needed by Q1 2027 · Open
- RQ-0144 · Credit card collections calls · Conversational audio · 5,000+ hours · needed by Q2 2027 · Matching
- RQ-0145 · Food-delivery order-issue calls with IVR handoff · Conversational audio · 3,000+ hours · needed by Q1 2027 · Open
- RQ-0146 · CT and MRI studies with radiology reports · Medical imaging · 50,000+ studies · needed by Q2 2027 · Open
- RQ-0147 · Mammograms with reports · Medical imaging · 20,000+ studies · needed by Q2 2027 · Open
- RQ-0148 · Chest X-rays with reports · Medical imaging · 100,000+ studies · needed by Q1 2027 · Matching

## Contact

- Labs and owners: https://quarry-website.vercel.app/request
- Email: hello@quarrydata.ai