Principal Researcher, AI Red Teaming (Remote, ITA)

Experteer Italy; Italia; Remote; EUR 120,000 - 180,000; Full time

Benefits offered by this job

Market-leading compensation & equity

Comprehensive wellness programs

Paid parental & adoption leaves

Professional development opportunities

Employee networks & volunteer opport.

Job summary

CrowdStrike seeks a Principal Researcher to set the strategic AI red-teaming agenda and perform the most challenging research in advancing its AI security platform.

You will lead frontier-model investigations, build repeatable tooling, and ensure findings are actionable across engineering, product, and legal for credible, public disclosures.

Qualifications

• Demonstrated hands‑on offensive security expertise (exploitation, vulnerability research, red teaming).
• Experience guiding researchers as a tech lead or principal.
• Exceptional written communication for external-facing notes and disclosures.

Responsibilities

• Define and prioritise attack classes, agent targets, and frontier-model behaviours with rationale and deadlines.
• Red team frontier models and agentic systems; track impact of model releases on attacks, behaviour, and controls.
• Conduct end-to-end vulnerability research and security assessments of agentic ecosystems.

Skills

Strong Python

Offensive security

Red teaming

Technical leadership

Written communication

Tools

Attack harnesses

Interceptors

Converters

Job description

As Principal Researcher, you set the strategic AI red-teaming agenda and execute the hardest research yourself to advance CrowdStrike’s AI security platform. You lead frontier-model investigations, build repeatable tooling, and raise the bar for peers, shaping disclosures and reusable attack modules. You work across engineering, product, and legal to ensure findings are actionable and publicly credible. This role combines deep technical leadership with high-impact, mission-driven research in AI safety.

• Market leader in compensation and equity awards
• Comprehensive physical and mental wellness programs
• Competitive vacation and holidays for recharge
• Paid parental and adoption leaves
• Professional development opportunities for all employees regardless of level or role
• Employee Networks, geographic neighborhood groups, and volunteer opportunities to build connections
• Define and prioritise attack classes, agent targets, and frontier-model behaviours with rationale and deadlines
• Red team frontier models and agentic systems; track impact of model releases on attacks, behaviour, and controls
• Conduct end-to-end vulnerability research and security assessments of agentic ecosystems
• Produce state-of-the-art research notes, external papers, talks, and coordinated disclosures
• Investigate failure modes in multi-agent systems and long‑horizon tasks
• Design and implement tooling for research and red‑teaming (attack harnesses, interceptors, converters)
• Convert findings into reusable attack modules for the automated AI security testing platform
• Provide technical guidance on external research agenda, disclosures, and collaborations
• Communicate findings and risk clearly to engineering, product owners, legal, and executives
• Review disclosures, attack classes, and lab designs as final technical reviewer
• Public record of original security research (CVEs, coordinated disclosures, talks, or papers)
• Deep hands‑on offensive security expertise (exploitation, vulnerability research, red teaming)
• Demonstrated research on LLMs or agentic systems (prompt/instrumentation techniques, multi‑agent attacks)
• Strong Python and systems engineering skills (harnesses, interceptors, converters)
• Understanding of agentic systems design (tool use, planning loops, memory, sandboxing, trust boundaries)
• Experience setting technical direction for researchers as a tech lead or principal
• Exceptional written communication for external-facing notes and disclosures

As Principal Researcher, you set the strategic AI red‑teaming agenda and execute the hardest research yourself to advance CrowdStrike’s AI s…