Jebra Research

Payroll rules are written in prose. Payroll systems need logic.

We're training a language model to turn natural-language pay agreements into structured, auditable payroll rules: the step before calculation.

Read the research Watch the demo

Pay agreement

studio_instructors_2026.pdf · §4 Rates

Instructors are paid $32 per class taught¹. Where a class has more than 20 attendees, each additional attendee adds $1.50². Classes on Sundays and public holidays are paid at time and a half³, and paid classes are capped at 12 per week⁴.

Extracted rule

schema v3

base_rate: 32.00 / class
attendance_bonus: 1.50 / head over 20
premium: ×1.5 · sun, public_holiday
weekly_cap: 12 classes

Needs reviewpublic_holiday: which calendar?

Where it fits

Jebra already runs the numbers. Now it's learning to read the rules.

Calculate · today

Jebra turns real shifts into pay: payslips, payouts and tax filings for shift-based businesses. It runs the numbers once the logic is known.

See how Jebra runs payroll

Interpret · in research

The step before calculation: reading pay agreements, contracts and notes, and turning them into that logic, so the rules Jebra pays by can come straight from the documents they're written in.

The problem

In shift-based work, the rules aren't structured yet.

Pay rules don't live in a database. They sit across pay agreements, contracts, spreadsheets, scheduling systems, manager notes and payroll history.

pay_agreement.pdf
contract.docx
rates_2026.xlsx
roster
manager_note.txt
payroll_history

Structured pay rules

A payroll manager can interpret that context. A payroll system can't use it, until it becomes structured logic.

The model

Trained for extraction, not conversation.

A language model fine-tuned on a single task: reading a pay agreement and producing a rule a payroll engine can execute.

01 Extract
Rates, units, premiums, caps and exceptions, pulled out of prose and named as fields.
02 Map
Pay terms mapped onto a structured payroll schema, so the same rule means the same thing every time.
03 Detect
Ambiguity flagged before a rule is used, not after it has already been paid.

Research

We publish the work as we do it.

Model performance on real and synthetic pay agreements, where off-the-shelf models fail, and how fine-tuning improves accuracy, including the results that don't go our way.

The interface

Ask your payroll anything.

A short walkthrough: ask any question about a pay period, in plain English.

Payroll rules are written in prose. Payroll systems need logic.

Jebra already runs the numbers. Now it's learning to read the rules.

In shift-based work, the rules aren't structured yet.

Trained for extraction, not conversation.

01 Extract

02 Map

03 Detect

We publish the work as we do it.

Fine-tuning with QLoRA for JSON extraction

Dicts and docs: the value of grammar and documentation for LLM-based automation

When payroll gets complicated, the machine gets harder to please

Ask your payroll anything.