$61–65/hr · Mercor · Part time, 7 hours a week
You evaluate whether AI systems understand advanced scientific topics correctly and handle sensitive information responsibly.
What you would do
- Write expert-level prompts in Danish across chemical, biological, and related scientific specialties
- Assess whether AI-generated responses show genuine understanding versus pattern-matching or hallucination
- Evaluate model performance on safety considerations around dual-use information and synthesis procedures
- Annotate responses using structured guidelines for accuracy, helpfulness, and responsible handling
Who they want
- PhD in chemistry, biology, or closely related field, whether ongoing or completed degree
- Native or near-native Danish fluency plus business-proficient written English for collaboration
- Deep technical familiarity with modern lab methods and computational approaches in your specialty
- Sound judgment about scientific safety, responsible disclosure, and hazardous-materials handling principles
- No AI or machine learning experience required; willingness to learn domain-specific evaluation workflow
Main skills
What the interview asks about
1.Dual-use recognition and judgment
AI systems can inadvertently help dangerous applications; your field-specific judgment prevents misuse while maintaining scientific openness.
For example: “You prompt the model to explain regioselectivity in an organic synthesis. The response is accurate but your domain knowledge suggests the pathway could enable synthesis of a controlled substance with minor modification. How would you annotate this response?”
2.Scientific accuracy in AI responses
Subtle errors in chemistry or biology can compound into dangerously wrong applications; catching AI mistakes requires research-level expertise.
For example: “The model explains a radiochemistry procedure using accurate foundational chemistry but incorrectly describes half-life decay consequences for shielding. Identify what the error is and why it matters for safety.”
3.Technical communication across languages
Danish technical vocabulary differs from English; you must ensure your prompts translate your intent faithfully and assess model understanding of Danish-specific terminology.
For example: “The Danish word for 'volatile' in chemistry has narrower meaning than English 'volatile.' Your prompt uses the Danish term but the model responds as if you meant the broader English definition. How would you recognize and flag this?”
4.Field-specific research responsibility
Different fields have different norms about disclosure; understanding what your discipline considers sensitive information versus routine knowledge shows safety maturity.
For example: “Biosafety Level 3 pathogen handling requires personal protective equipment per your national guidelines. If the model recommends Level 2 protocols for a Level 3 agent, explain why this is dangerous and how you'd annotate it.”
A task you may get
Write three Danish prompts at foundational, advanced, and dual-use levels. For each, predict what accurate responses should include and evaluate whether a sample response covers safety considerations.
How to prepare
- List five chemistry or biology topics where you've published or conducted research, then note which touch on dual-use, hazardous materials, or restricted procedures
- Review your field's guidelines on responsible disclosure and safety-restricted information, noting what's freely published versus access-controlled
- Prepare examples of AI errors you've observed outside your field to calibrate your expectations for where models commonly fail on technical topics
The facts
- Pay
- $61–65/hr
- Hours
- Part time, 7 hours a week
- Where
- Remote · Remote — Western Europe 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.