EXPERT HUMAN DATA FOR AI
Training examples, human feedback and expert evaluations.
Built for AI teams. Grounded in European context.
THE IDEA IS SIMPLE
Some knowledge comes from doing. The judgment behind a difficult decision. The detail a specialist knows to look for.
Legendre brings domain experts together to create and review data for AI. You define what your model needs to learn. We organise the people, the work and the delivery.
See how it worksEXPERTISE, APPLIED
Training and evaluation data built around the decisions your model needs to get right.
Assess a financial model. Explain which assumptions support the conclusion, and where the reasoning falls short.
The judgment to spot inconsistencies, challenge assumptions and describe a well-supported answer.
Expert-written tasks · Reference answers · Review rubrics
Compare two contract clauses. Explain the material differences, using only the supplied text.
The judgment to account for jurisdiction, catch ambiguity and flag conclusions the source cannot support.
Comparison tasks · Annotated answers · Review criteria
Investigate a software issue. Compare possible fixes, account for the constraints and explain how to test the result.
Practical reasoning about trade-offs, failure modes and the evidence needed to choose a reliable next step.
Problem sets · Worked solutions · Expert feedback
Review a model’s response in a chosen language. Judge whether it preserves the meaning and fits its audience and setting.
A feel for terminology, tone and cultural nuance, with explanations of why one response works better than another.
Language tasks · Response comparisons · Review notes
Training examples, human feedback and expert evaluation.
The expertise and deliverables are agreed for each project.
EUROPE IS THE STARTING POINT
Fluency is only part of a good answer. Terminology, professional practice and local expectations matter too.
Our focus is expert data for European contexts. We define the language, expertise and data requirements with your team before work begins.
Discuss your use caseThe right terminology, nuance and tone.
The details a practitioner would check.
The assumptions a generic answer can miss.
DEFINED IN THE BRIEF. APPLIED IN THE REVIEW.
HOW IT WORKS
A paid pilot. A clear standard.
A dataset your team can assess.
Agree the task, domain, languages and acceptance criteria. Set the budget and data-handling terms before production.
A shared briefIdentify the expertise your project needs. Calibrate contributors with sample tasks, clear instructions and a common review rubric.
An aligned expert teamReview the work against the agreed criteria and resolve disagreements. Deliver the accepted data in the format your team needs.
Reviewed project dataYour team pays Legendre for the project. We coordinate contributors and delivery.
FIELD NOTES
PRACTICAL GUIDE · 5 MIN READ
Six decisions to make before the first task is written.
Read the guideA FEW USEFUL DETAILS
AI labs and teams that need specialist human input to train, refine or evaluate their models. We focus on projects where domain knowledge, language and context matter.
The scope can include expert sourcing, contributor calibration, task creation, human feedback, review and data delivery. The brief specifies the work, acceptance criteria and commercial terms.
Yes. A focused pilot is the starting point we propose: one use case, a defined set of tasks and a clear way to judge the result before expanding.
We scope language coverage, expert qualifications, access, storage and usage rights with your team before work begins. These requirements are agreed for each project.
LET’S START A CONVERSATION
Tell us what you’re working on.
A few sentences are enough to start.