AI security engineer

We are seeking an AI Security Engineer to evaluate the cybersecurity capabilities and safety of frontier AI models. The role will focus on developing internal cyber benchmarks and test how well model guardrails refuse cyber-related requests. The work will inform how the firm assesses and safely adopts AI models across the enterprise.

• Design, build, and maintain internal benchmarks that measure frontier models’ cyber capabilities.
• Develop and run test suites that measure refusal rates for policy-violating cyber requests, including adversarial and jailbreak-style prompts.
• Maintain evaluation datasets, scoring methods, and tooling so tests are repeatable across model versions and vendors.
• Partner with internal teams and stakeholders, such as Model Risk Management, governance, and technology teams to understand evaluation needs and support safe AI adoption.
• Communicate results clearly to both technical and non-technical audiences through reports, dashboards, and presentations, including actionable recommendations.
• A suite of internal cyber benchmarks that measure frontier model capabilities across defined security tasks, with documented methodology and scoring criteria.
• A repeatable guardrail testing framework that measures how models respond to cyber tasks across defined risk categories, confirming they refuse or comply appropriately for each.
• A maintained library of evaluation datasets and test prompts mapped to each risk category, versioned so results can be compared over time.
• Supporting tooling and documentation that allow teams to run evaluations independently.

• Bachelor's or Master's degree in Computer Science, Software Engineering, Cybersecurity, Data Science, Artificial Intelligence, or a related field.
• Будет плюсом: Hands-on cybersecurity experience in areas such as offensive security, penetration testing, red teaming or vulnerability research; Proficiency in Python or another programming language for building test harnesses, automating tasks, and analyzing results; Strong written communication skills for documenting methods and findings; Experience using and prompting large language models, exposure to model evaluation or AI safety testing, or a strong curiosity and ability to quickly learn how these models work and how to assess them; Experience working with APIs and building automation around them

• Cooperation on a B2B contract basis.
• Opportunity to use foreign languages and the newest technologies on a daily basis.