Adversarial Prompt Expert

Handshake; San Francisco (CA); Remote; USD 150,000 - 200,000; Part time

Job summary

A technology company is seeking LLM testers to engage in a red teaming project focused on evaluating large language models. The role involves crafting prompts, documenting outcomes, and working independently. Ideal candidates are doctoral students or recent graduates with practical LLM experience and a strong creative mindset. This is a part-time, flexible remote position.

Qualifications

• Hands-on experience with multiple LLM models.
• Skill in crafting prompts and evasion techniques.
• Willingness to think creatively and push boundaries.

Responsibilities

• Work independently on a red teaming project.
• Develop prompts and evaluate LLM responses.
• Document outcomes clearly and systematically.

Skills

LLM Usage

Prompt Engineering

Adversarial Mindset

Creativity

Documentation Skills

Ethical Awareness

Education

Doctoral students or recent graduates

Job description

Overview

You’ll be part of a red teaming project focused on probing large language models for failure modes and harmful outputs. Your work will involve crafting prompts and scenarios to test model guardrails, exploring creative ways to bypass restrictions, and systematically documenting outcomes. You’ll think like an adversary to uncover weaknesses, while collaborating with engineers and safety researchers to share findings and improve system defenses.

Responsibilities

• Remote and asynchronous work; work independently from anywhere.
• Flexible hours and approximately 10 to 20 hours per week.
• Develop domain-specific prompts and evaluate LLM responses as part of project work.
• Dedicate time researching topics of interest with AI assistance.
• Learn new skills while contributing to AI across disciplines.
• Placement into a project will be dependent on project availability.

Qualifications

• Heavy LLM Usage — hands-on experience with multiple models (open- and closed-source), comfort experimenting across systems.
• Prompt Engineering & Jailbreaking — skill in crafting prompts, evasion techniques, and creative ways to bypass restrictions.
• Adversarial / Security Mindset — ability to think like an attacker, with bonus points for red teaming or offensive security background.
• Persistence & Creativity — willingness to try many variations, think outside the box, and push edge cases.
• Clear Documentation — ability to log attempts and outcomes systematically, and communicate issues clearly.
• Ethical Awareness — understands boundaries and handles sensitive content responsibly.

This program is open to U.S.-based doctoral students, candidates, and recent graduates with valid work or training authorization (e.g., F-1/OPT, J-1, H-1B). Participants are responsible for ensuring compliance with their visa conditions and confirming eligibility with their program or visa sponsor prior to applying.

• At this time, we are unable to accommodate candidates on STEM OPT who require an i-983. Fellows with already approved i-983s, as well as those on pre-grad OPT, CPT, J-1, or H-1B, are not impacted. This position may be subject to change in the future.