AI Jailbreak Detection Researcher

Yovex Intelligence

Remote · $130k – $160k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Adversarial prompt attacks against LLMs are not a research curiosity — they're an active threat to every customer-facing AI deployment. Our red team evaluates AI systems for enterprise clients and government agencies, and jailbreak detection is one of the most requested capabilities. The problem isn't just detecting known attack patterns: it's detecting novel attacks that don't match any signature we've seen before. You'll lead the research and engineering for our jailbreak detection system: studying the latest attack methodologies, building classifiers that generalize beyond training distribution, developing adversarial robustness testing frameworks, and publishing findings that help the broader safety community. The dual mandate — detection and publication — means you need to be both a rigorous researcher and a practical engineer who ships systems clients can deploy. We work with Fortune 500 companies whose AI systems face sophisticated adversarial users. The attacks we see in production are often more creative than what's in the literature. You'll have access to a proprietary dataset of 2 million real-world attack attempts that no academic lab has.

Fully remote.

Responsibilities

Design and train jailbreak detection classifiers generalizing beyond known attack signatures
Build adversarial robustness evaluation frameworks for client AI deployments
Analyze novel attack patterns from production dataset and distill into model improvements
Develop transfer attack simulations for stress-testing deployed detection systems
Publish research findings on adversarial robustness and jailbreak detection
Collaborate with client security teams on deployment and threshold calibration

Requirements

6+ years in ML research with adversarial ML, AI safety, or NLP security focus
Deep understanding of LLM attack taxonomy: prompt injection, jailbreaking, extraction
Strong Python and PyTorch skills with adversarial training experience
Publication record in adversarial ML, AI safety, or related security venues
Ability to work under responsible disclosure protocols