Research Scientist/Engineer (Agentic Systems)

About This Opportunity

We are representing a well-funded AI safety organization in Paris that is building cutting-edge systems to test, evaluate, and improve the reliability of large language model agents. Our client works at the frontier of agentic AI research, focusing on understanding how autonomous systems fail and how to prevent those failures before deployment.

We're seeking a Research Scientist/Engineer to join their fundamental research team. You'll design and build large-scale adversarial environments that systematically push AI agents to their breaking points, studying failure modes that emerge in both single-agent and multi-agent settings.

Responsibilities

• Design and implement complex adversarial environments that expose the boundary between agent capability and alignment, creating scenarios where failures are informative rather than trivial
• Build realistic multi-agent systems with comprehensive instrumentation to observe emergent breakdowns and unscripted failure patterns
• Run end-to-end experiments at scale, orchestrating parallel agent systems against both external APIs and proprietary models
• Systematically catalogue concrete agent failure modes and develop tooling to surface these failures across large-scale deployments
• Translate experimental findings into both internal behavioral models and public research outputs

What We're Looking For

Required:

• Demonstrated experience building at least one non-trivial agent environment or automated research pipeline that ran end to end, with clear understanding of failure analysis
• Strong software and AI engineering capabilities with proven ability to independently orchestrate many agents and containers in parallel
• Track record of empirical research in agents, red-teaming, or post-training where you defined research questions, executed experiments, and drew defensible conclusions
• Fast empirical iteration skills with comfort working in ambiguous problem spaces—ability to transform fuzzy concerns into concrete, falsifiable experiments
• Daily fluency with frontier language models and coding agents as a power user

Strong advantages:

• Published research at top-tier venues on automated red-teaming, agentic environments, or post-training
• Experience building monitoring systems for model failures and anomalous behavior detection
• Track record reproducing public benchmark results and identifying methodological fragility
• MSc or PhD in machine learning, computer science, cognitive science, computational neuroscience, physics, or related quantitative field
• Participation in AI safety fellowships (MATS, ASTRA, Anthropic Fellows) or comparable self-directed research

How We Work With You

When you apply through Get A Job.ai, our talent team will review your profile and conduct an initial screening. If there's a strong match, we'll submit your application to our client on a confidential basis. The client's process includes a take-home test task, technical interview with their Head of Fundamental Research, and final conversation with leadership.

Important: Please do not attempt to identify or contact the client directly. All communication regarding this role should go through Get A Job.ai to protect both your candidacy and our client relationship.

The role offers hybrid work in Paris with relocation support available, or remote work from London (note: relocation support and medical insurance are not available for London-based candidates). Benefits include competitive equity, comprehensive medical coverage for France-based team members, all necessary hardware and tools, covered AI tool subscriptions, and biannual team off-sites.

Pay

Compensation details will be discussed during the interview process based on experience and location.

Equal Opportunity: Get A Job.ai is committed to inclusive recruiting practices. We partner with clients who value diverse perspectives and welcome candidates from all backgrounds.