$53–57/hr · Mercor · Part time, 7 hours a week
A PhD-level scientist fluent in Czech and English who evaluates AI model safety by crafting prompts and assessing responses across specialized STEM domains.
What you would do
- Compose technically sophisticated prompts in Czech addressing chemistry, biology, radiological or nuclear subjects to test model understanding
- Review machine-generated responses for scientific precision, relevance to modern laboratory practice, and computational accuracy
- Judge whether outputs handle information about hazardous materials, biosecurity, or dual-use capabilities with appropriate responsibility
- Apply standardized classification systems to categorize model behavior against safety benchmarks
- Provide structured feedback documenting assessment reasoning and safety concerns
Who they want
- PhD holder in chemistry, biology, molecular science, radiochemistry, or adjacent disciplines, completed or in progress
- Native or fluent Czech speaker with professional business English proficiency in both written and spoken forms
- Hands-on laboratory or computational research experience demonstrating deep familiarity with methods and equipment in your subfield
- Sound professional judgment regarding chemical safety protocols, biosafety procedures, and responsible handling of sensitive technical information
- Part-time availability for contract work at 53-57 USD hourly, immediate start date possible
Main skills
What the interview asks about
1.Prompt construction across specialized domains
Interviewers assess your ability to generate technically rigorous queries that test model competency specifically, requiring both subject mastery and understanding of what gaps matter in your field.
For example: “Write a Czech prompt about organic synthesis protocols that would effectively test whether a model understands proper safety precautions. What specific details would you include to make it scientifically meaningful?”
2.Evaluating factual accuracy in technical output
Your credibility depends on catching subtle errors or omissions that reflect incomplete domain knowledge, distinguishing between minor lapses and serious misstatements.
For example: “A model describes a molecular weight calculation method in Czech with standard terminology but applies an outdated formula from a 1990s reference. How would you assess the response, and what guidance would you offer?”
3.Judging sensitive information disclosure appropriately
This role centers on recognizing when models reveal dual-use details irresponsibly, requiring research experience and discretion about what constitutes reasonable risk.
For example: “A response covers radiochemical synthesis and includes unnecessary procedural detail about handling materials. How would you weigh scientific completeness against safety concerns in your assessment?”
4.Applying standardized safety classification systems
Consistency in categorizing model behavior across many examples ensures reliable patterns and actionable feedback for model improvement.
For example: “You've encountered five similar prompts about biochemical pathways with varying model responses. Describe how you would categorize them against guidelines to highlight systematic strengths or weaknesses.”
5.Bridging Czech and English technical communication
Precision in conveying complex scientific concepts across languages prevents misunderstanding and ensures assessments reflect actual model behavior rather than translation artifacts.
For example: “A Czech prompt uses a colloquial chemistry term that has no direct English equivalent. How would you document the prompt's intent for non-Czech speakers reviewing your assessment?”
How to prepare
- Prepare a written overview of your dissertation or primary research focus, highlighting key methods and the most hazardous materials or techniques your work involved
- List five current chemical or biological topics you consider sensitive in terms of dual-use risks, and your reasoning for each classification
- Review any laboratory safety certifications or biosafety protocols you have completed, and be ready to discuss specific situations where you applied disciplinary judgment
The facts
- Pay
- $53–57/hr
- Hours
- Part time, 7 hours a week
- Where
- Remote · Remote — Eastern 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.