$30–34/hr · Mercor · Part time, 7 hours a week
Write advanced Turkish-language science prompts and evaluate how AI responds to them for accuracy and safety.
What you would do
- Compose challenging scientific prompts in Turkish spanning chemistry, biology, and nuclear/radiological fields
- Evaluate model responses for scientific accuracy, depth, and appropriate handling of dual-use concerns
- Classify prompts and conversations using structured guidelines and taxonomies
- Judge whether model recommendations are responsible given the sensitivity of the topic
- Provide feedback indicating where model responses succeed or miss the mark
Who they want
- PhD in chemistry, biology, or closely related field (completed or in progress)
- Native or near-native Turkish fluency and professional-level written English
- Deep familiarity with modern lab techniques and computational methods in your specialty
- Sound judgment about scientific safety and dual-use research implications
- Preferably: experience with radiochemistry, nuclear chemistry, or synthesis topics
Main skills
What the interview asks about
1.Prompt design at expert depth
A vague prompt wastes evaluation time; your ability to write scientifically rigorous questions tests whether you truly understand your field.
For example: “Write a Turkish synthesis prompt testing green chemistry understanding and safety evaluation. Why would this reveal model reasoning?”
2.Dual-use judgment and sensitivity
Your assessment of what constitutes a safety concern shapes whether the model learns to be responsible.
For example: “Model describes accurate purification methods applicable to sensitive synthesis. How do you rate technical correctness vs. safety risk?”
3.Evaluation against scientific norms
Accuracy means more than getting facts right; it includes whether a response reflects the real state of knowledge in your field.
For example: “A model responds to a chemistry prompt by citing a procedure from a 1990 paper. The technique works, but better methods exist that are safer and more efficient. How would you rate the response, and what would you note about its limitations?”
4.Translation and terminology precision
Turkish scientific terminology has specific meanings; an answer can be technically wrong through imprecise translation or terminology confusion.
For example: “You read a model response in Turkish that uses a term that could mean either compound X or compound Y depending on context. The response is clear in English, but ambiguous in Turkish. How does this affect your assessment?”
How to prepare
- Refresh your knowledge of dual-use research concerns in your field and how scientific communication can be responsible without blocking legitimate research
- Prepare 3-4 scientifically challenging prompts in Turkish that span different subfields and would test AI understanding
- Review how scientific accuracy in model responses should be evaluated: technical correctness, depth, completeness, responsible framing
- Study examples of how models handle sensitive topics and practice identifying where they succeed or fall short
The facts
- Pay
- $30–34/hr
- Hours
- Part time, 7 hours a week
- Where
- Remote · Remote — MENA 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.