Remote Applied Mathematician - AI Reasoning Trainer

DataAnnotation; Canada; Remote; CAD 79.000 - 245.000; Vollzeit

Benefits dieser Stelle

Choose projects that you work on

Work from home on your computer

Pay rate $40-$125 USD per hour withBon

Zusammenfassung

DataAnnotation is looking for an experienced mathematician to design mathematical problems that today's leading AI models still get wrong. Models are fluent enough to produce an answer that looks right and a derivation that reads well.

You bring the judgment to tell the difference between the two, and the tooling to prove it. This is a new and growing area for our company, focused on teaching AI to reason correctly rather than plausibly.

Qualifikationen

• Fluency in English (native or bilingual level).
• Active research background in mathematics or statistics: advanced graduate student, postdoc, academic researcher, or industry practitioner.
• Strong coding skills. You write Python and work comfortably at a terminal as part of your research.
• Familiarity with reproducible or containerized environments such as Docker is a plus.

Aufgaben

• Design complex mathematical tasks in your area of expertise.
• Define what a correct answer looks like and the standard a model has to meet.
• Test and validate the task yourself before you submit it.
• Write a step-by-step solution another researcher could reproduce.
• Design tasks a model cannot solve by replaying a known result, and flag where its reasoning breaks: invalid steps, unstated assumptions, right answers reached by wrong work.
• Work across numerical analysis, optimization, probability and statistics, applied analysis, and scientific machine learning.

Kenntnisse

Fluency in English

Python programming

Ausbildung

Active research background in mathematics or statistics

Tools

Docker

Jobbeschreibung

DataAnnotation is looking for an experienced mathematician to design mathematical problems that today's leading AI models still get wrong. Models are fluent enough to produce an answer that looks right and a derivation that reads well.

You bring the judgment to tell the difference between the two, and the tooling to prove it. This is a new and growing area for our company, focused on teaching AI to reason correctly rather than plausibly.