Adversarial Task Writer (AI)

TL;DR

Adversarial Task Writer (AI): Designing and implementing prompt injection scenarios in YAML to test the robustness of frontier AI models with an accent on adversarial mindset and security policy validation. Focus on building realistic simulated environments and executing systematic red teaming tests to identify agent vulnerabilities.

Location: Must be based in Serbia, Armenia, or Bulgaria

Company

A specialized provider of evaluation infrastructure for AI safety and robustness testing, simulating adversarial conditions for AI development teams.

What you will do

• Design prompt injection scenarios in YAML to test AI agent safety policies.
• Simulate realistic environments such as e-commerce, finance, or enterprise SaaS platforms.
• Develop adversarial payloads embedded in messages, tool responses, and documents.
• Execute systematic tests against frontier models using Docker and CLI tools.
• Validate success rates and ensure tasks meet quality gates for policy violations.
• Maintain high-quality output standards for adversarial task submissions.

Requirements

• Must be based in Serbia, Armenia, or Bulgaria
• Expertise in direct and indirect prompt injection techniques.
• Proficiency in YAML for technical writing and scenario design.
• Experience with Docker and CLI tools for systematic testing.
• Strong background in pentesting, AppSec, or LLM security research.
• Domain expertise in at least one vertical like finance, healthcare, or e-commerce.

Culture & Benefits

• Opportunity to work on cutting-edge AI safety and red teaming infrastructure.
• Flexible remote work environment within specified locations.
• Performance-based compensation model per accepted task.
• Engagement with advanced AI security research and frontier model testing.