AI Red Team – GenAI Adversarial Evaluation

Location: Ramat Gan, Israel; onsite, full-time

Company

hirify.global is a trust, safety, and security company focused on AI security, model evaluations, red-teaming, runtime guardrails, and drift detection.

What you will do

• Review adversarial AI tasks for realistic threat models, plausible attacker capabilities, reachable poisoned content, and realistic target systems.
• Assess whether reported harm is genuinely caused by the adversary rather than by a rigged or preconfigured task.
• Validate success and failure criteria, reward logic, deterministic checks, and resistance to reward hacks or evasion.
• Measure ASR@K, variance, reproducibility, attack novelty, sophistication, stealth, and coverage across harm types and target surfaces.
• Identify anomalies and systemic issues across large task sets and define validation guidelines, quality thresholds, and release gates.
• Build automated validation, regression, and quality-monitoring pipelines while collaborating with red-teamers, researchers, and task authors.

Requirements

• 4+ years of experience in analytical, ML, security research, or applied research roles, including hands-on adversarial ML, LLM security, red-teaming, or AI safety evaluation.
• Bachelor’s or Master’s degree in Computer Science, Machine Learning, Statistics, a related quantitative field, or equivalent practical experience.
• Working knowledge of prompt injection, indirect prompt injection, jailbreaks, data exfiltration, or tool-misuse attacks against LLM systems or agents.
• Strong Python and data-analysis skills, with grounding in statistics, experiment design, variance analysis, hypothesis testing, and anomaly detection.
• Experience building or auditing data-driven validation frameworks, using LLMs for analysis or evaluation, and creating automated tests.
• Strong adversarial thinking, reproducibility expertise, and excellent analytical, problem-solving, and communication skills.

Nice to have

• Offensive cybersecurity certifications such as OSWA, OSWE, OSCE3, SEC542, or SEC522.
• Experience designing or evaluating AI agents, LLM-powered applications, agentic workflows, or adversarial benchmark datasets.
• Familiarity with RL Gym-style task formats, Gymnasium, OpenAI Gym, or agent-evaluation harnesses.
• Experience with automated evaluation pipelines, regression testing, quality gates, safety/security taxonomies, or threat modeling.

Culture & Benefits

• Work on AI trust, safety, and security for frontier model labs, enterprises, and user-generated-content platforms.
• Contribute to systems supporting the safety of AI and communication technologies used by more than 3 billion people.
• Collaborate with red-teamers, researchers, and task authors on improving adversarial evaluation quality.