Cyber Security Engineer (Offensive Security)

At Sabio Group, we build and operate AI-powered customer experience platforms for some of the world's most demanding enterprise brands. As we push deeper into agentic AI, multi-cloud architectures and high-volume conversational systems, the attack surface evolves with us and we need people who can think like adversaries to keep our customers, our data and our reputation safe.

About the Role

We're hiring a Red Team Security Engineer to join our Information Security & Cyber Security team in South Africa. You'll be the offensive counterpart to our defensive function: stress-testing the solutions we design, build and operate from cloud-native services to LLM-powered agents by attacking them the way a real adversary would, and partnering with engineering teams to fix what you find.

This is a hands-on role for someone who is genuinely curious about how systems break, comfortable writing code as well as reading it, and excited about the new class of vulnerabilities emerging in AI and agentic systems.

Key Responsibilities

Offensive Security & Red Teaming

• Plan and execute red team engagements, penetration tests and adversary simulations against our platforms, products and corporate environment.
• Identify, exploit and document vulnerabilities across web applications, APIs, cloud infrastructure, identity systems and AI/LLM-based services.
• Develop realistic attack scenarios — initial access, privilege escalation, lateral movement, data exfiltration — mapped to frameworks such as MITRE ATT&CK.
• Build and maintain custom tooling, exploits and automation where off-the-shelf tools fall short.
• Leverage AI pen testing tooling and frameworks as a force amplifier within your role.

AI Powered Security Tooling & Frameworks

• Active, hands-on use of AI-powered offensive security tooling as a core part of your workflow — leveraging LLMs and agentic assistants to accelerate reconnaissance, exploit development, code review, payload generation and report writing.
• Familiarity with emerging AI red-team frameworks and platforms — e.g. PyRIT, Garak, Promptfoo, NVIDIA Aegis, Microsoft Counterfit, HackerOne / Bugcrowd AI testing toolkits, or equivalent — and a practical sense of when to use which.
• Experience building or extending automated AI red-team harnesses : prompt-injection test suites, jailbreak corpora, tool-abuse scenarios, multi-turn attack agents, and regression eval sets for LLM and agentic systems.
• Pragmatic judgement on the limits and risks of AI-assisted offensive work — hallucinated findings, false confidence, data leakage into third-party models — and the discipline to validate AI output before acting on it.
• Curiosity to keep pace with a fast-moving space: new models, new attack techniques, new defensive controls — and a willingness to bring those learnings back into the team.

AI & Agentic System Security

• Probe LLM-powered and agentic systems for prompt injection, jailbreaks, tool/function-call abuse, context leakage, insecure output handling and supply-chain risks.
• Test RAG pipelines, knowledge bases and integrations for data exfiltration, poisoning and unauthorised access paths.
• Contribute to internal threat models for agentic architectures and help shape secure-by-default patterns for multi-agent and tool-using systems.
• Stay close to evolving standards and guidance (e.g. OWASP Top 10 for LLMs, NIST AI RMF, emerging agent interoperability protocols).

Cloud & Application Security Testing

• Perform deep-dive security testing of cloud workloads across AWS, Azure and/or GCP — IAM, network, container, serverless and data-layer concerns.
• Review CI/CD pipelines, IaC (Terraform, CloudFormation, Bicep) and Kubernetes deployments for misconfigurations and exploitable weaknesses.
• Conduct source-assisted ("grey-box") testing — reading code to find logic flaws, authZ gaps and unsafe integrations.

Vulnerability Management & Remediation Partnership

• Triage findings, assign realistic severity, and write clear, reproducible reports with concrete remediation guidance.
• Partner with engineering teams to validate fixes, advise on secure design and pair on hardening work — not just throw findings over the wall.
• Drive continuous improvement of detection coverage by working with the blue team / SOC on purple-team exercises.

Tooling, Automation & Continuous Testing

• Build automation that turns one-off tests into repeatable, scheduled checks — exposure scanning, attack-path analysis, agent red-teaming harnesses.
• Integrate offensive testing into the SDLC: SAST/DAST/IAST, dependency scanning, secrets detection, container and IaC scanning.
• Treat evaluation and regression of security controls as a first-class deliverable — measured, not assumed.

Collaboration & Responsible Disclosure

• Work alongside the Head of Information Security, AI Ethics leads, platform engineering and product teams to embed security early.
• Produce clear design reviews, threat models, runbooks and post-engagement reports for both technical and executive audiences.
• Operate within strict rules of engagement, with care for production stability, customer data and legal/regulatory obligations.

Skills Knowledge and Expertise

Required

• Demonstrable hands-on experience in offensive security — penetration testing, red teaming, or adversary simulation — against modern web, API and cloud-based systems.
• Strong understanding of common vulnerability classes (OWASP Top 10, authZ flaws, SSRF, deserialisation, injection, cryptographic misuse) and how to chain them into real impact.
• Solid grasp of cloud security in at least one major provider (AWS, Azure or GCP): IAM, networking, key management, container and serverless services, common misconfigurations and attack paths.
• Practical understanding of AI/LLM systems — how they work, where they fail, and the new risks they introduce (prompt injection, jailbreaks, insecure tool use, training/RAG data exposure).
• Coding capability in at least one of Python, Go, JavaScript/TypeScript or similar — comfortable writing exploits, tooling and automation, not just running other people's tools.
• Confidence with offensive tooling — Burp Suite, nmap, Nuclei, BloodHound, Metasploit, custom scripts — and the judgement to know when to build vs. buy.
• Familiarity with CI/CD, containers and IaC (Docker, Kubernetes, Terraform or equivalent) and how to attack and defend them.
• An evaluation mindset : you measure security posture with reproducible tests, metrics and evidence — not gut feel.
• Comfort with agentic development workflows — using AI coding assistants and AI co-work / pair-development models (Claude Code, Copilot, Cursor or equivalent) as part of your day-to-day delivery.
• Clear written and verbal communication in English: able to brief engineers, executives and (where relevant) customers on findings and risk.
• A strong ethical compass and discipline around scope, rules of engagement, evidence handling and responsible disclosure.

Desirable

• Industry certifications such as OSCP, OSEP, OSWE, CRTO, CRTP, GPEN, GXPN, GCPN, AWS/Azure/GCP security specialties or equivalent.
• Experience red-teaming or evaluating agentic AI / LLM systems in production — prompt injection campaigns, tool-abuse testing, multi-agent attack scenarios, AI red-team frameworks.
• Exposure to purple teaming and detection engineering: working with SOC/SIEM (e.g. Sentinel, Splunk, Defender XDR) to improve detections from offensive findings.
• Experience with identity attacks across Entra ID / Azure AD, Active Directory, OAuth/OIDC and federated environments.
• Source-code review s