AI Security Researcher/Research Engineer
Research and engineer security measures for LLM-powered web agents and chatbots, including adversarial testing, secure architectures, model evaluation, and security-focused training. The role requires strong Python and ML expertise, AI security experience, and preferably a computer science Ph.D.
About the job
Responsibilities
- Design and execute adversarial evaluations to uncover failure modes in LLM-powered web agents and chatbots, including prompt injection, tool abuse, and data exfiltration.
- Analyze model behavior, including internal thoughts, to understand and forecast security risks in frontier models.
- Develop and prototype security-focused training and fine-tuning techniques.
- Augment web agents with security principles such as information-flow control, privilege control, and contextual security.
- Partner with engineering, privacy, machine learning, and cryptographic researchers to translate findings into product changes, tests, and measurable security improvements.
- Help define security strategy and policy for building and responsibly releasing web agents, with a focus on user protection.
Requirements
- Demonstrated experience with secure AI, including prompt-injection attacks and security guardrails.
- Strong publication record in AI security, ML security, or closely related areas.
- Track record of shipping products and features in a fast-moving environment.
- Strong Python coding skills and familiarity with ML frameworks such as PyTorch or JAX.
- Experience training, fine-tuning, and evaluating large language models.
- Deep familiarity with large language models and web agents, including how they work and are trained.
- Experience applying security principles such as information-flow control and access control/least privilege.
- Demonstrated ability to bring clarity and ownership to ambiguous technical problems.
- Ph.D. in computer science highly preferred; exceptions may be considered for candidates with industry experience.
Nice-to-haves
- Knowledge of AI-system pipelines and security evaluation.
- Open-source projects focused on security or privacy.
- Experience with confidential computing and ZKP/SMPC protocols.
- Ph.D. in computer science from a top-tier school.
- Strong publication record in premier systems, performance, or machine-learning venues.
- Industry work experience.
Skills
Python, PyTorch, JAX, LLMs, Web Agents, Prompt Injection, Security Guardrails, Fine-Tuning, Information Flow Control, Access Control, Confidential Computing, Zero-Knowledge Proofs, Smpc, Adversarial Evaluation, Privacy
Similar jobs
AI Research jobs
Improbable
Remote
Conduct applied research on AI agents, designing experiments and evaluation systems to improve reliability, context retention, and multi-step task completion. The role requires strong AI/ML research, engineering, experimental design, and communication skills.
Brave
London, United Kingdom
Research and engineer AI systems, optimizing and evaluating machine-learning models and on-device performance across desktop and mobile products. The role requires software development experience, systems and networking knowledge, and expertise with modern ML frameworks, privacy, and security.
Mercor
San Francisco, CAResearch fellows propose, build, validate, and publish benchmarks or evaluation methodologies for measuring frontier AI performance on economically valuable professional and scientific work. The fellowship requires a specific research pitch, relevant technical or adjacent-field background, and a commitment of at least 20 hours per week.
PostHog
United Kingdom
Researches and productionizes machine-learning models for session replay analysis, user behavior prediction, and synthetic testing. The role requires strong mathematics, PyTorch, systems programming in Rust/CUDA/C, and a product-oriented approach to shipping.
AI Digest
Remote
Conduct research on long-horizon, multi-agent AI behavior by designing agent environments, analyzing large-scale data, and running experiments. The role requires strong research judgment, rapid execution, independence, and familiarity with current AI developments.