STEM PhDs and Technical Domain Experts

What you would do

• Design difficult, realistic scenarios that stretch model capabilities
• Curate datasets representing complexity within your discipline
• Evaluate data quality and consistency across deliverables
• Document technical rationale and problem selection logic

Who they want

• PhD or equivalent experience in STEM fields like medicine, statistics, computer science, or engineering
• Demonstrated ability to communicate technical concepts with exceptional clarity
• Track record of meticulous attention to detail in technical work
• Self-directed professional comfortable working independently

Main skills

What the interview asks about

1. 1.Domain expertise application

   The role demands you recognize subtle, complex scenarios that surface model weaknesses; interviewers verify you grasp your field's depth.

   For example: “Describe a counterintuitive problem in your domain where naive models fail. Why would engineers training systems need examples like this?”

2. 2.Problem design rigor

   Creating effective training problems requires balancing realism with challenge; this separates candidates who superficially understand their field.

   For example: “Walk me through how you'd create five diverse, difficult problems for a dataset on medical diagnostics. How would you ensure they represent genuine clinical complexity?”

3. 3.Technical communication

   You'll explain domain nuance to researchers unfamiliar with your specialty; miscommunication wastes training resources and degrades data quality.

   For example: “Imagine an AI researcher outside your field reads your problem description. What three specific details would you include to prevent misunderstanding?”

4. 4.Quality consistency

   Without direct supervision, you maintain standards across hundreds of examples; this separates reliable contributors from unreliable ones.

   For example: “You've created 200 problems and notice 10% have subtle inconsistencies in difficulty or documentation. How would you fix and prevent this?”

5. 5.Edge case identification

   Models learn from edge cases; you must instinctively recognize when a scenario is unusual, rare, or challenges assumptions.

   For example: “You're designing dataset problems for computer vision. Which edge cases in your domain would you prioritize, and why would they reveal model gaps?”

A task you may get

Create 3-5 technical problems in your domain that are difficult but realistic, with clear documentation of why each challenges current approaches.

How to prepare

• Review recent papers or research in your field highlighting open challenges or model limitations
• Prepare 2-3 examples of subtle complexities in your domain that non-specialists often miss
• Practice explaining one complex concept from your field in plain language
• Reflect on past work where attention to detail or rigor directly impacted outcomes

The facts

Pay

$55–80/hr

Hours

Hourly

Where

Remote

Open to

CAN, USA, GBR, AUS

Field

Data Analysis

Posted

5/19/2026

Places left

104

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