Path 8 of 10, by 2026 demand
Security Engineer (AI and cloud)
Secure AI features and the cloud platforms under them: threat models, identity, data protection, agent permissions and detection.
Demand The fastest-rising niche in this list: AI Security Engineer demand up 268% in 2026 while overall US cyber postings fell 26% (third-party, M5); AI/ML and cybersecurity tie as the hardest IT roles to fill (third-party, M6). Kept at #8 for fit with your background, not for demand.
What the role is
AI and cloud security engineers design and review controls for systems that use models and agents: prompt-injection resistance, least-privilege tools, data isolation, identity and secrets, supply chain, and monitoring for abuse. Most also own classic cloud security: landing zones, network segmentation, workload identity, policy as code.
Both labs hire security engineers in application security, detection and response, and AI-specific areas (our count, §2.3).
Level signals
- Threat models that lead to concrete, prioritised controls.
- Security as an enabler: paved roads, not just reviews.
- Depth in identity and cloud networking.
- Clear risk communication to engineers and executives.
The typical loop
No single public source. Common shape for security engineering roles.
| Stage | What it tests | Basis |
|---|---|---|
| Coding | Scripting and secure code review | Common industry pattern; no single source |
| Security design / threat model | An AI feature or cloud platform | Common industry pattern; no single source |
| Scenario / incident | Detection and response walk-through | Common industry pattern; no single source |
| Cloud depth | Identity, networking, policy | Common industry pattern; no single source |
| Behavioural | Influence, risk trade-offs | Common industry pattern; no single source |
Track order
Work top to bottom. Ticks come from the lessons you mark complete.
- 01Securing AI systemsSecurity engineering for AI systems track. The core of the role.
- 02Building blocksLLM and agent engineering track. Understand what you are securing.
- 03Production AI systemsLLM and agent engineering track. Guardrails and agent harnesses.
- 04The trade-offsSystem design track. Reliability and isolation patterns.
- 05Worked promptsSystem design track. Enterprise Q&A permissions deep dive.
- 06Arrays and stringsCoding patterns track. Scripting fluency.
- 07Behavioural rounds at Staff levelLeadership and behavioral track. Influence and pushing back stories.
- 08Mock roundsMock rounds track. Design and behavioural mocks.
Role practice set
Prompts written for this site. They are not reported interview questions.
- Threat-model an agent that reads email, browses the web and can send replies; produce the tool policy table.
- Design tenant isolation for a vector index with regulated customers.
- Review a code sample where model output is inserted into SQL and HTML; list and fix the issues.
- Explain to an executive in two minutes why prompt injection is an architecture problem.
Your résumé and story angle
- Architect-grade security: identity, network segmentation and data residency on Azure are daily work for an enterprise architect. [Security review or landing zone you designed.]
- Kubernetes hardening: [pod security, network policy, workload identity work].
- AI security: [any AI integration where you designed controls]. If thin, build the practice 1 threat model as a portfolio piece.
- Gap to close: detection and response, offensive testing. Be explicit about which side you are pitching.