Director, AI Safety and Evaluation Centre of Excellence Lead

Some careers have more impact than others.
If you’re looking for a career where you can make a real impression, join HSBC and discover how valued you’ll be.

We are currently seeking an experienced professional to join our team in the role of POSITION TITLE.

Business: Chief AI Office

Job ID：56199

Role Description:

Joining HSBC’s Chief AI Office (CAIO) places you at the forefront of Artificial Intelligence (AI) in one of the world’s leading financial institutions. In a highly collaborative and fast-paced environment, you will be part of a team responsible for HSBC’s group-wide AI strategy and driving the adoption of AI at scale.

Responsible AI at HSBC is actionable, risk-based, and outcome-oriented—bridging the philosophical and the practical by translating principles into concrete processes, controls, and accountability structures. Reporting to the Group Head of Responsible AI, the AI Safety and Evaluation Centre of Excellence (COE) Lead will manage a small, agile team of experts developing internal standards, shape industry best practice, and coordinate applied AI safety testing in partnership with AI leaders across the Group. You will play a critical role in embedding enhanced safety and evaluation practices into the AI lifecycle, ensuring that AI is deployed responsibly and aligned with regulatory and organisational expectations.

The role demands deep expertise in AI safety and evaluation techniques, a strong understanding of operating models for AI governance, and the ability to translate complex concepts into practical implementation. To succeed, you will bring a combination of subject matter expertise at the cutting edge of AI development and risk management, alongside exceptional communication skills, critical thinking, and the ability to influence at senior levels across a global financial institution.

Responsibilities:

The successful candidate will be empowered to lead the development of an enterprise capability for AI Safety and Evaluation, aligning to the Group AI strategy and Responsible AI objectives. Expected responsibilities include:

AI Safety Testing & Evaluation

• Build and lead a high-performing AI Safety and Evaluation CoE, establishing scalable testing capabilities, reusable assets, and best practices, including but not limited to ‘orange teaming’, simulation, and automated evaluation complimenting existing Cyber and MRM testing.
• Test and challenge guardrails, testing and evaluation methodologies for use cases, informing governance and approval decisions. Including:
  • Business risk translation & governance: Translate results into clear business risk readable assessments (customer harm, conduct, financial crime, legal/regulatory, reputational, operational risk), with impact/likelihood scoring, acceptance criteria, and decision-ready go/no-go recommendations.
  • End-to-end control effectiveness testing: Evaluate not just model quality but the full solution (human-in-the-loop, monitoring, escalation, fallback/kill-switch), including control design and effectiveness evidence for approvals/audit.
  • GenAI system security & agentic controls: Test for prompt injection, tool/retrieval poisoning, unsafe tool use, least-privilege permissions, action confirmation/transaction boundaries, and broader supply-chain/configuration risks.
  • Data, privacy & information governance: Assess data provenance/lineage and usage rights (training/fine-tuning/RAG), cross-border/retention constraints, and run privacy testing (PII leakage, memorisation/regurgitation, membership inference) with mitigations.
  • Operational resilience & reliability: Define and test non-functional requirements (availability, latency, dependency/provider outage scenarios, safe degradation), plus ongoing monitoring for drift, safety regression, and incident triggers.
• Risk-tiered metrics & automation: Define KRIs/KPIs and regression benchmarks and integrate evaluation into LLMOps/CI/CD as automated test harnesses and release gates with traceable reporting.
• Provide senior subject matter expertise, collaborating with stakeholders across the CAIO, Risk, Cyber, and AI leadership across the Group to strengthen AI safety and responsible AI adoption.

Standards, Frameworks & Operating Model

• Define, implement, and continuously improve internal standards, methodologies, and tooling for scaled AI safety and evaluation testing and assurance across the AI lifecycle.
• Establish and operationalise a consistent target operating model for AI safety and evaluation testing, complimentary to existing Model Risk Management (MRM), Cyber Security, and Business Risk frameworks.
• Translate responsible AI principles into practical controls, test protocols, and measurable assurance criteria embedded within governance processes.

External Engagement & Thought Leadership

• Assess emerging AI risks (e.g., misuse, adversarial threats, and agentic behaviours).
• Monitor and interpret developments across industry, translating insights into HSBC’s AI safety and evaluation approach.
• Represent HSBC in external forums and working groups, contributing to the advancement of industry best practice in AI safety and evaluation.

Requirements:

Minimum Requirements

AI Engineering & Systems

• Strong working knowledge of AI/ML systems, including Generative AI/LLMs and deployment patterns (e.g. APIs, pipelines, MLOps/LLMOps).
• Experience embedding guardrails and safety controls into model development and deployment lifecycles.

Model Evaluation & Testing

• Hands-on experience with model evaluation techniques (performance, bias/fairness, explainability, robustness).
• Practical experience conducting AI safety testing (e.g. red teaming, adversarial testing, scenario-based evaluation).

Cybersecurity & AI Threats

• Understanding of AI-related cyber risks (e.g. prompt injection, data leakage, adversarial attacks) and threat modelling concepts.
• Ability to work with Cyber Security teams to integrate threat-led testing into evaluation processes.

Risk & Governance

• Familiarity with Model Risk Management (MRM) and operational risk frameworks in regulated environments.
• Ability to design and document testing outcomes, link to risk decisions, and support governance approvals.
• Experience in the application of key frameworks (e.g. NIST AI RMF, ISO/IEC 42001, MAESTRO).

Preferred Requirements

Advanced AI Engineering & Tooling

• Deep expertise in LLM/agentic architectures, evaluation frameworks, and alignment techniques.
• Experience with modern tooling such as Python ML stack (PyTorch, TensorFlow), evaluation frameworks (e.g. LangChain, prompt/eval harnesses), and scalable testing infrastructure.

Advanced Evaluation & Red Teaming

• Proven experience designing enterprise-scale AI red teaming programmes, including automated evaluation, simulation, and synthetic data generation.
• Expertise in defining risk-tiered evaluation strategies and metrics aligned to business criticality and regulatory expectations.
• Deep practical understanding of adversarial AI security practices (e.g. model inversion, data poisoning, jailbreak techniques) and integration into enterprise controls.
• Experience aligning AI testing with broader cyber assurance and penetration testing programmes.

Risk Leadership & Operating Model Design

• Experience designing and scaling AI safety or model validation capabilities within financial services or similarly regulated industries.
• Strong alignment with MRM, audit, and regulatory expectations, including documentation, traceability, and audit readiness.

Industry Experience

• Significant experience in financial services, or regulated industries deploying AI at scale.
• Engagement with regulators, industry bodies, and academic communities on AI safety, assurance, or governance topics.