AI Security Researcher

Overview

AI systems are being deployed at pace into environments where failure carries consequence. Enovise's AI Security Research function exists to understand, test, and defend against the misuse and compromise of AI and autonomous systems — before threat actors and adversarial states exploit them. As an AI Security Researcher, you will investigate adversarial machine learning, prompt injection, model theft, agentic system abuse, and emerging AI-enabled attack vectors. Your research feeds directly into client engagements, product development at our innovation platforms, and Enovise's threat intelligence capability. This is research with operational consequence.

Responsibilities

• Research adversarial attacks against large language models, vision models, and autonomous agentic systems.
• Design and execute adversarial testing methodologies for AI systems deployed in government and enterprise environments.
• Investigate and document attack vectors including prompt injection, jailbreaking, model inversion, membership inference, and data poisoning.
• Produce research output — papers, briefings, and internal reports — that translates findings into practical defensive recommendations.
• Advise on AI system procurement, deployment architecture, and governance from a security perspective.
• Collaborate with Redpherix product teams to integrate research findings into autonomous red teaming capabilities.
• Track emerging academic and operational research in AI security and synthesise relevance for client risk.

Requirements

• Deep understanding of machine learning fundamentals — supervised, unsupervised, and reinforcement learning — and modern neural network architectures.
• Demonstrated experience researching or testing AI system vulnerabilities in a professional or academic context.
• Proficiency in Python and familiarity with ML frameworks (PyTorch, TensorFlow, or Hugging Face).
• Strong grasp of adversarial ML concepts: evasion, poisoning, extraction, and inference attacks.
• Ability to communicate research findings clearly to both technical and non-technical audiences.
• Eligibility for security clearance — cleared work experience is highly advantageous.
• Published research, CVEs, or responsible disclosure in AI/ML security is a strong differentiator.

Nice to Have

• Experience red-teaming LLM-based applications and agentic AI pipelines.
• Understanding of AI governance frameworks (NIST AI RMF, EU AI Act) and their security implications.
• Background in formal verification, model robustness research, or AI safety.
• Experience with autonomous systems in defence or critical infrastructure contexts.