STEM Data Annotation Specialist

Frontier models are now measured substantially on their reasoning in mathematics, physics, chemistry, biology, engineering and computer science. Improving that requires people who can tell a correct derivation from a plausible-looking wrong one, which is a much narrower population than general annotation requires and is paid accordingly.

You will write problems, evaluate model reasoning step by step, and correct solutions in your discipline. Fully remote across Canada, minimum twenty hours weekly. This work suits graduate students, postdoctoral researchers, working engineers and educators who want substantive part-time work.

Evaluate model reasoning in your discipline, identifying not only whether the answer is wrong but where in the chain the reasoning failed, which is the part that actually improves a model. Author original problems and reference solutions at specified difficulty levels. Correct and rewrite model-generated solutions with a clear rationale for each change. Assess code correctness, efficiency and edge case handling on software engineering programmes. Verify claims against authoritative sources rather than intuition. Document ambiguity and contribute to guideline refinement where the current version produces inconsistent outcomes.

A degree in a STEM discipline, and in most programmes graduate level study or equivalent professional depth. Genuine subject expertise in at least one of: pure or applied mathematics, statistics, physics, chemistry, biology, computer science, software engineering, electrical, mechanical, civil or chemical engineering. Ability to work with LaTeX or comparable notation. English at C1. Rigour, and the intellectual honesty to mark something wrong when you are not certain rather than waving it through.

Graduate degree or doctorate. Teaching, tutoring, marking or peer review experience. Publication record. Competitive mathematics or programming background. French at C1, which opens bilingual STEM programmes that pay at a premium. Professional engineering designation.

Rates well above general annotation, reflecting the scarcity of the expertise. Fully remote and genuinely flexible, which works alongside graduate study or a full-time technical role. Paid guideline training. Work that materially and measurably improves model reasoning in your field. Routes into senior review and programme design.