Annotation & labelling
Label text and organise examples using guidelines agreed with your team. Arabic and English are our focus.
AI annotation & RLHF
We organise data labelling, human feedback and response evaluation for AI teams. Projects are scoped around the task, language and subject expertise required.
Discuss a projectWhat we can help with
Human feedback for RLHF
Preference comparisons and response rankings to support your training pipeline.
Model evaluation
Review outputs against agreed rubrics for relevance, accuracy and task performance.
Specialist review
Identify the language and subject expertise needed, with reviewer availability confirmed during scoping.
How the project works
Scope the task
Define guidelines, deliverables, reviewer requirements and quality targets.
Build the workflow
Confirm reviewer coverage, tools and access arrangements.
Calibrate with a pilot
Use a small batch to refine the rubric and resolve disagreements.
Review and deliver
Deliver agreed batches with quality review and feedback.
Common questions.
Do you run the full RLHF training process?
This service focuses on the human-data layer: annotation, preference ranking and response evaluation. Model training infrastructure is outside the standard scope.
Which languages and dialects can you support?
Arabic and English are the focus. Specific dialects and reviewer coverage are confirmed before a project starts.
Can a project start with a pilot?
Yes. A defined pilot helps establish the guidelines, workflow and acceptance criteria before a larger engagement.