Extracting data from driving course certificates
A course provider issues a certificate when a driver finishes a defensive driving course, and each state sets its own name and format. Insurers check one before granting a discount, and traffic courts before dismissing a ticket. One arrives for every driver claiming the discount, as an emailed PDF, portal upload or printed copy.
An underwriter needs the driver's name and license number, the certificate number, the course type and the completion date. A policy administration system applies the discount from that date and sets its expiry. A court's case management system records the completion and dismisses the charge.
What makes driving course certificates hard
Every value prints twice
Uniform certificates such as the Texas form print a court copy above an insurance copy on one page, giving each value two candidate locations.
One course goes by many names
Texas calls it a driving safety course and New York a point and insurance reduction course, and each provider designs its own certificate.
Eligibility turns on date and purpose
The completion date starts the discount period, and Florida's statute excludes a course that a court imposed for a moving violation.
Parse, extract, verify
Parse
Parse turns the document into reading-order Markdown and a hierarchy of blocks (text, tables, figures), and grounds each block to its exact location on the page.
Extract
You define the fields in a JSON schema, and Extract returns their values in a consistent, structured format, each linked back to where it appears in the Markdown.
Verify
Each value links back to where it appears, down to the character range in the Markdown, and the page and bounding box in the source document, so every result can be verified.
One driving course certificate, end to end
About this document: A State of Texas driving safety course uniform certificate of completion, issued by a course provider that the Texas Department of Licensing and Regulation approves. One portrait page prints the certificate twice, a court copy above an insurance copy, with a large agency watermark behind both. Page 1 is featured. The student's name, address, date of birth, license number and certificate number were replaced with stand-ins.
Processing objective: The objective is to read the court copy as one complete filing, with each value grounded to the top copy and kept apart from its duplicate below. The course was completed to dismiss a citation, so this copy also carries a court block that the insurance copy leaves out.
Demonstration fields: Certificate number, Court, Reason for course, Completion date, Issue date
The document
1 page
Page 1shown below
Where each value came from
certificate_number
{
"value": "CP275-63203533",
"page": 1,
"range": {
"end": 1185,
"start": 1157
},
"box": {
"xmin": 0.72387,
"ymin": 0.03148,
"xmax": 0.95311,
"ymax": 0.04437
}
}court
{
"value": "Ellis County: Municipal Garrett (Garrett)",
"page": 1,
"range": {
"end": 900,
"start": 858
},
"box": {
"xmin": 0.09276,
"ymin": 0.28839,
"xmax": 0.36007,
"ymax": 0.30201
}
}reason
{
"value": "Traffic Ticket",
"page": 1,
"range": {
"end": 1545,
"start": 1530
},
"box": {
"xmin": 0.7322,
"ymin": 0.44547,
"xmax": 0.81337,
"ymax": 0.45741
}
}completion_date
Printed on the page as 01/24/2022, normalized on extraction.
{
"value": "2022-01-24",
"page": 1,
"range": {
"end": 1470,
"start": 1459
},
"box": {
"xmin": 0.72694,
"ymin": 0.32284,
"xmax": 0.80329,
"ymax": 0.33556
}
}issue_date
Printed on the page as 01/24/2022, normalized on extraction.
{
"value": "2022-01-24",
"page": 1,
"range": {
"end": 1495,
"start": 1484
},
"box": {
"xmin": 0.72778,
"ymin": 0.36421,
"xmax": 0.80346,
"ymax": 0.37643
}
}How to try it yourself
Everything on this page is reproducible. The sample, its schema, and the Parse and Extract API outputs are published. You can repeat the same two calls from whichever surface you already build on.
- Claude Skilldescribe the job; the agent writes it
- CLIno code at all
- MCPexpose ADE docs as an agent tool
- REST APIany language
- Pythonlandingai-ade
- TypeScriptlandingai-ade
Documents into data
Accurate
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Auditable
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Agentic
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
API-first
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Affordable
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Enterprise-grade
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Built for high-volume, high-throughput production
- Zero data retention
- Priority rate limits
- Enhanced support
- VPC and on-prem deployment
Available on Team and Enterprise plans. VPC and on-prem deployment require an Enterprise plan.
