Extracting data from employment offer letters
Every hire starts with an offer letter, and no two employers write one the same way. What counts as compensation differs by country: an Indian letter breaks out HRA and provident fund, a US one states base and bonus. Onboarding and background-check teams handle thousands, as emailed PDFs and scans of the signed copy.
Onboarding needs the designation, the start date, the place of posting and the compensation. An HR information system opens the employee record from those, payroll seeds the salary structure, and an immigration filing quotes the title and the figure.
What makes employment offer letters hard
The terms are sentences
Designation, joining date and pay are each stated in prose: a sentence beginning "Your Date of Joining will be" is the whole field.
Salary structures have no standard
An Indian letter itemizes HRA, provident fund and professional tax. A letter from another country uses an entirely local vocabulary for the same idea.
Quantities written as words
Probation runs "Six months", retirement comes at "Fifty-eight years", and pay is given in figures and again in words.
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 employment offer letter, end to end
About this document: An offer letter from Aerial Telecom Solutions, an Indian telecom services company, for a Technical Manager post. Five portrait pages exported from Word, with a logo, a watermark and a branded footer throughout. Pages 1 to 4 are the letter and its terms; page 5 is Annexure I, the salary breakdown. This is a real letter, and the candidate's name, address and the signatory's signature were removed before publication.
Processing objective: The objective is to lift the terms of the offer from the letter that states them. Each of the five values sits inside a sentence of running prose. The box covers that sentence, so a reviewer reads the commitment in the employer's own words.
Demonstration fields: Designation, Place of posting, Date of joining, Annual cost to company (Rs.), Letter date
The document
4 of 5 pages
Page 1shown below
Page 2
Page 3
Page 5
Where each value came from
offer.designation
{
"value": "Technical Manager",
"page": 1,
"range": {
"end": 390,
"start": 300
},
"box": {
"xmin": 0.07867,
"ymin": 0.42146,
"xmax": 0.72851,
"ymax": 0.43759
}
}offer.place_of_posting
{
"value": "Jalandhar",
"page": 1,
"range": {
"end": 701,
"start": 653
},
"box": {
"xmin": 0.07839,
"ymin": 0.56679,
"xmax": 0.43017,
"ymax": 0.58559
}
}offer.date_of_joining
{
"value": "07-Apr-2025",
"page": 1,
"range": {
"end": 744,
"start": 703
},
"box": {
"xmin": 0.0787,
"ymin": 0.60157,
"xmax": 0.38177,
"ymax": 0.61734
}
}offer.annual_ctc
{
"value": 565008,
"page": 1,
"range": {
"end": 976,
"start": 842
},
"box": {
"xmin": 0.07634,
"ymin": 0.70109,
"xmax": 0.88886,
"ymax": 0.73538
}
}letter_date
{
"value": "20-Mar-2025",
"page": 1,
"range": {
"end": 65,
"start": 53
},
"box": {
"xmin": 0.77902,
"ymin": 0.12629,
"xmax": 0.87747,
"ymax": 0.13747
}
}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.
