English Writer - Freelance AI Trainer (Creative & Adversarial)

Mindrift connects specialists with leading tech companies to test and improve AI systems. They are looking for writers who can think like "Controlled Adversaries." Your job is to engage in realistic, multi-turn conversations with Large Language Models (LLMs) to stress-test them. You will adopt various personas to expose the model's weaknesses while maintaining narrative coherence.

Key Responsibilities

• Roleplay & Persona Adoption: Put yourself in the position of different user personas and engage in realistic dialogue, using a range of tones and registers.
• Stress-Testing: Deliberately engineer conversational pressure to test the model's emotional intelligence, instruction-following, and consistency.
• Human Simulation: Mimic how humans actually speak—including hesitations, contradictions, and emotional escalation—to see how the AI reacts.
• Analysis & Documentation: Pinpoint failure modes in the LLM's responses and document your observations clearly.

Requirements

• Language Proficiency: C2-level English (Native or near-native proficiency). Examples: CPE, TOEFL 114+, IELTS 8.0+.
• Background:
• Undergraduate or Postgraduate qualification in an Arts-based subject (English, Creative Writing, Journalism, Psychology, Linguistics) OR
• 1+ years of experience in Conversational AI Testing, Narrative Design, or Adversarial Model Testing.
• Mindset: You must be able to think like both a Storyteller and a Tester.

Nice to Have

• Background in Screenwriting, Interactive Fiction, or Narrative Design.
• Understanding of Prompt Engineering or LLM evaluation concepts.
• Experience with behavioral research or linguistics.

How it Works

1. Apply with your resume (in English).
2. Pass Qualification tests (Language & Logic).
3. Join a Project and receive training/mentorship.
4. Complete Tasks (approx 10-20 hours/week).
5. Get Paid.