- Level
- Senior
- Experience
- Not stated
- Work style
- Remote
- Full-time
- Area
- Offensive security
At a glance
The role in plain English
This role focuses on offensive security research targeting AI systems, particularly large language models and autonomous agents. It suits experienced security professionals with a strong background in red teaming, penetration testing, and experimental design. Day-to-day work includes developing attack tooling, conducting adversarial evaluations, analysing failures, and collaborating with technical teams to turn research findings into practical security improvements.
What you'll do
- Research and exploit weaknesses in LLM applications and agents
- Develop attack tooling and reproducible evaluation harnesses
- Design and run experimental attack scenarios and benchmarks
- Analyse failures and validate exploits with actionable reporting
- Collaborate with teams to translate findings into security improvements
What they're looking for
Essential
- Experience in AI red teaming or offensive security
- Strong Python and research engineering skills
- Ability to design and run controlled experiments
- Clear written and verbal communication skills
Nice to have
- Experience with agent orchestration or harness engineering
- Knowledge of agent security or guardrail bypass
- Familiarity with evals and benchmarks for LLMs
Core skills
- AI red teaming
- Penetration testing
- Python
- Experimental design
- Failure analysis
- Agent orchestration
The shape of the job
Technical depth
4/5
Pace
3/5
Ownership
4/5
Teamwork
4/5
Cyberend's reading of the listing, not the employer's own scores.
Application toolkit
Make your application count
See which requirements your CV meets and what is missing.
Your CV is used for this answer only and never stored.
Questions to prepare for
- 01Describe a prompt injection attack you have researched or mitigated.
- 02How would you design an evaluation harness for LLM agent security?
- 03Explain your approach to analysing failures in AI-driven systems.
- 04How do you communicate complex security findings to non-technical stakeholders?
About Snyk
Snyk develops security platforms and products, with a focus on improving software and AI security.
Full job descriptionAs published by Snyk
Unleash AI Innovators, Securely
Software is being rewritten in real time. AI agents are starting to write more code than humans do, at a speed no security team has ever had to match, and that single shift is opening up the biggest opportunity our industry has seen in a decade. We're not bracing for it, we're building the company that owns it.
Snyk is building the next generation security platform where security isn't a set of tools people operate. It's a team of security agents working alongside your engineers, finding, fixing, and governing risk at the speed software is now built. These agents work alongside both the company build and the agents they adopt. Security stops being a queue of work waiting for people, and becomes a coordinated workforce companies direct and trust.
Research and exploit weaknesses in LLM applications and autonomous agents. Build attack tooling, agent harnesses, and reproducible evaluations that turn offensive research into measurable improvements in AI security and improve on Snyk’s offensive security product offering.
Looking for candidates with experience in the following areas:
- AI red teaming: Prompt injection, jailbreaks, data leakage, tool abuse, and multi-step adversarial attacks.
- Offensive security: Web/API penetration testing, authentication and authorization flaws, business logic attacks, and exploit chaining.
- LLM fundamentals: Model behavior and limitations, inference, context management, retrieval, and tool calling.
- Harness engineering: Agent orchestration, tool integrations, state management, sandboxed execution, and reproducible runs.
- Agent security: Trust boundaries, permissions, memory poisoning, scope enforcement, and guardrail bypass testing.
- Evals and benchmarks: Realistic attack scenarios, ground truth, reliable graders, dataset quality, and regression suites.
- Experimental design: Controlled comparisons, repeated trials, attack success rates, precision/recall, and cost/performance measurement.
- Research engineering: Strong Python, APIs, containers, browser automation, and building maintainable research tools.
- Failure analysis: Inspecting prompts, traces, and tool outputs to distinguish model, harness, and environment failures.
- Exploit validation and reporting: Reproducible proofs of concept, demonstrated impact, false-positive analysis, and actionable remediation.
Collaboration and communication:
- Collaborative research: Experience delivering research in a team and translating findings into actionable recommendations that shape priorities and strategy.
- Written and verbal communication: Clearly communicate complex research findings, risks, and tradeoffs to technical and non-technical stakeholders.
About Snyk
Snyk is committed to creating an inclusive and engaging environment where our employees can thrive as we rally behind our common mission to make the digital world a safer place. From Snyk employee resource groups, to global benefits that help our employees prioritize their health, wellness, financial security, and a work/life blend, we aim to support our employees along their entire journeys here at Snyk.
Equal Employment Opportunity
Snyk is an Equal Opportunity Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, or status as a protected veteran.