Job Description

Location: London / Hybrid - 2/3 days onsite
Sector: UK Financial Services
Rate: Competitive Market Rates (Role Dependent)
IR35: Inside
Day rate: up to 600 per day
Length: 9 months

All candidates must demonstrate:

• UK FinServ Compliance: Deep familiarity with DORA, FCA Operational Resilience, and the EU AI Act.

• AWS Bedrock Stack: Hands-on experience with Agents, Knowledge Bases, Guardrails, and model lifecycles.

• AI/ML Core: Mastery of FMs, RAG, tool-use, and non-deterministic agent logic.

• Security Frameworks: Proficiency in OWASP LLM Top 10 and NIST AI RMF.

• Mission: Secure the "Human-to-Agent" and "Agent-to-Resource" identity chain.

• Key Tasks: Lead SPIFFE/SPIRE rollouts with AWS Roles Anywhere; harden OBO (On-Behalf-Of) flows.

• Security: Implement sender-constrained tokens (PoP) and JIT for non-human identities.

• Threats: Modelling for Confused Deputy and Federation Hijacks; ensuring full traceability per EU AI Act Art. 12 & 14.

• Mission: Lead structured offensive modeling to identify systemic AI vulnerabilities.

• Key Tasks: Conduct STRIDE modeling and build attack trees for prompt injection and "Denial-of-Wallet" scenarios.

• Validation: Run scenario workshops to test real-world control effectiveness.

• Governance: Maintain an adversarial knowledge base and map safeguards to compliance frameworks.

• Mission: Automate the "Release Gate" for AI systems through rigorous adversarial testing.

• Key Tasks: Integrate tools like Pyrit, Garak, or AgentDojo into CI/CD pipelines.

• Metrics: Build frameworks for uncertainty, drift detection, and success rates.

• Assurance: Own the AI-BOM and generate Article 15 evidence for bias, hallucination, and memorisation risks

Randstad Technologies is acting as an Employment Business in relation to this vacancy.