Training Turk

Bilingual Norwegian STEM Expert (PhD) — AI Safety

$77–81/hr · Mercor · Part time, 7 hours a week

A PhD scientist with native Norwegian fluency evaluates how AI models handle specialized chemistry and biology knowledge.

What you would do

  • Craft technical prompts in Norwegian across chemistry, biology, and nuclear domains to rigorously test model capabilities and identify limitations.
  • Evaluate model responses for scientific accuracy, handling of sensitive topics, and appropriateness regarding dual-use information.
  • Apply structured frameworks to systematically classify scientific conversations and assess content against established safety guidelines.
  • Leverage specialized research knowledge to verify whether model outputs demonstrate genuine domain understanding.

Who they want

  • Doctoral degree in chemistry, biology, or closely related scientific field, whether current or completed.
  • Native or near-native Norwegian language skill combined with business-level English writing proficiency.
  • Deep knowledge of laboratory techniques and computational methods within your specialized research area.
  • Sound judgment regarding scientific safety, biosafety procedures, and responsible treatment of hazardous material topics.

What the interview asks about

  1. 1.Norwegian scientific prompt creation

    Building prompts that rigorously test model knowledge in Norwegian is your core function; interviewers assess depth across multiple domains.

    For example: “If you needed to thoroughly probe a model's understanding of nuclear chemistry, what sequence of Norwegian prompts would you develop to move from basic concepts through advanced edge cases involving material hazards?”

  2. 2.Dual-use risk discernment

    Distinguishing between legitimate scientific inquiry and information with misuse potential is essential for model safety; this drives hiring decisions.

    For example: “A model responds in Norwegian describing synthesis procedures for a controlled chemical. How would you assess whether the specificity level raises concerns, and what framework guides your determination?”

  3. 3.Technical expertise demonstration

    Evaluating model accuracy depends on genuine subject matter mastery; interviewers probe whether your background gives you real credibility.

    For example: “A model asserts that compound X undergoes a particular reaction under given conditions with product Z. How would you validate this claim and what experimental knowledge informs your assessment?”

  4. 4.Norwegian-English technical precision

    Accurately creating and evaluating scientific Norwegian requires language fluency and domain mastery together; interviewers test both.

    For example: “Explain a concept like chemical equilibrium to a non-specialist first in Norwegian, then in English. What terminology challenges emerge, and how do you preserve precision in both languages?”

  5. 5.Systematic evaluation and documentation

    Consistent judgment across many evaluations requires careful rubric adherence; interviewers assess your ability to apply standards methodically.

    For example: “You receive a classification framework with five sensitivity levels. One prompt seems to fit levels 2 and 4 depending on interpretation. How do you reach a defensible judgment and document your reasoning?”

A task you may get

Given a set of Norwegian chemistry prompts and corresponding model responses, classify each on a sensitivity scale, identify scientific errors, and assess whether the model handles dual-use knowledge appropriately.

How to prepare

  • Study your most specialized research area thoroughly and prepare to critically evaluate AI claims within that domain.
  • Review dual-use considerations in chemistry and biology to develop a framework distinguishing appropriate from problematic scientific discussion.
  • Practice composing technical explanations in Norwegian and English on your research work to strengthen bilingual clarity and precision.
  • Explore how evaluation rubrics function in content safety and model assessment frameworks.

The facts

Pay
$77–81/hr
Hours
Part time, 7 hours a week
Where
Remote · Remote — Norway preferred
Posted
9/4/2026

We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.