Senior AI Security Researcher
Lead research into the security of frontier AI models and agentic systems, discovering novel attack techniques and building rigorous methods to evaluate and mitigate them.
About the role
We are looking for an exceptional security or machine learning researcher who wants to work on problems that do not yet have established answers. You will investigate how advanced AI systems behave under adversarial conditions, develop new attack techniques and evaluation methodologies, and work closely with engineering teams to turn research into production-grade security capabilities. The work spans LLMs, multimodal models, AI agents, tool use, prompt injection, model manipulation, data poisoning, information leakage, jailbreaks, insecure model integrations, and emerging threats created by increasingly autonomous systems. You will have significant freedom to identify important problems, design experiments, build prototypes, and determine where our research should go next. This is a hands-on research role. We expect you to write code, build attack harnesses, analyze model behavior, reproduce results, and communicate your findings to both technical and non-technical audiences.
Responsibilities
- Identify and investigate previously poorly understood security failure modes in LLMs, multimodal models, and agentic AI systems.
- Develop novel attack techniques, adversarial examples, exploit chains, and proof-of-concept demonstrations.
- Design rigorous evaluations for model robustness, agent security, prompt injection resistance, data leakage, and tool-use safety.
- Build automated red-team infrastructure capable of testing models and AI applications at scale.
- Analyze model and agent behavior using quantitative experiments, statistical analysis, and reproducible research methodologies.
- Develop threat models for AI systems and translate research findings into concrete security controls.
- Collaborate with ML engineers and product teams to turn research prototypes into reliable production capabilities.
- Track emerging research and independently reproduce important results from academic papers and industry work.
- Publish or otherwise communicate high-quality research when appropriate, including technical reports, internal research notes, and conference-quality work.
- Present findings clearly to engineering teams, security practitioners, customers, and company leadership.
- Mentor other researchers and raise the technical quality of the broader research organization.
Requirements
- Several years of experience in security research, machine learning research, applied AI, offensive security, or a closely related technical field.
- Demonstrated ability to independently identify an ambiguous technical problem, formulate hypotheses, design experiments, and produce a convincing result.
- Strong understanding of modern LLM architectures, model behavior, inference, fine-tuning, or evaluation techniques.
- Hands-on experience with adversarial ML, AI red teaming, application security, offensive security, or security research.
- Strong Python programming skills and the ability to build research tooling from scratch.
- Experience working with ML frameworks and tooling such as PyTorch, Hugging Face Transformers, or equivalent technologies.
- Strong understanding of security fundamentals including threat modeling, attack surfaces, authentication, authorization, data exposure, and application security.
- Ability to read and critically evaluate current AI/security research papers.
- Strong written and verbal communication skills.
- Evidence of exceptional technical ability through research publications, open-source projects, security research, CTFs, bug bounties, shipped systems, or equivalent work.