Assistant Director, AI Enablement & Security Engineering

Position Summary

The Team Lead, AI Enablement & Security Engineering is a hands-on technical leadership role responsible for enabling the secure adoption of Generative AI, Large Language Models and agentic AI solutions across the IT division.

The role will go beyond policy writing to translate AI security, governance and risk requirements into practical architectures, engineering controls, reusable development patterns and assurance processes. The incumbent will design, prototype, implement and assess safeguards for LLM applications, Retrieval-Augmented Generation pipelines, AI agents and supporting AI platform services.

Working closely with cybersecurity, cloud platform, enterprise architecture, data governance, risk and application teams, the Team Lead will establish secure-by-design practices that allow AI solutions to be deployed safely, efficiently and at scale. The role will also build internal engineering capability through technical guidance, reusable assets, hands-on workshops and mentoring.

Key Responsibilities

AI Security Engineering and Technical Governance
Define and maintain secure-by-design architecture patterns, engineering standards, technical controls and security baselines for Generative AI, RAG and agentic AI solutions, aligned with OWASP, NIST AI RMF, MITRE ATLAS and applicable Singapore AI security guidance.
Design, implement and evaluate AI runtime guardrails across inputs, outputs, tool calls and data access to mitigate prompt injection, sensitive-data exposure, harmful content, excessive agency, insecure output handling and other adversarial behaviours.
Lead AI threat modelling, security assessments, testing and red-teaming across models, application, prompts, RAG pipelines, agent workflows, APIs, plugins, MCP servers and supporting infrastructure.
Design secure RAG architectures and AI agents incorporating identity-aware retrieval, access controls, metadata filtering, tenant isolation, source validation, retrieval-poisoning protection, least-privilege access, tool allowlisting, execution boundaries, approval checkpoints, credential isolation and auditable human oversight.
Conduct technical assurance reviews of internal and third-party AI solutions by validating architectures, configurations, vendor claims, security evidence and controlled proof-of-concept outcomes.
Contribute technical inputs to the AI inventory, covering deployed models, owners, approved data sources, hosting locations, dependencies, security classifications and control status.
Partner with cybersecurity, risk, legal, data protection and governance team to translate policies and risk decisions and assurance outcomes into enforceable platform and application controls.
Hands-on AI Enablement, Architecture and Prototyping
Design and prototype secure, scalable AI solutions using commercial and open-source models, cloud AI services, AI gateways, orchestration frameworks and agent development platforms.
Develop reference architectures and implementation patterns for common use cases such as enterprise chat, knowledge assistants, secure RAG, workflow automation, coding assistants and tool-using agents.
Implement or configure AI gateway controls covering identity, authentication, authorisation, model access, secret management, quotas, rate limits, content inspection, routing, fallback, logging and cost controls.
Create reusable, pre-hardened development assets, including code templates, deployment pipelines, prompt patterns, agent configurations, security test cases, policy-as-code controls and infrastructure-as-code modules.
Establish secure AI development and deployment practices into DevSecOps pipelines, architecture reviews, release gates and production-readiness assessments.
Evaluate emerging AI models, frameworks, agent protocols, guardrail technologies and security tools through structured experiments, proofs of concept and applied technical learning
Provide hands-on technical advisory and troubleshooting support to project teams during solution design, prototyping, security review and production onboarding.
Balance rapid experimentation with enterprise requirements for security, data protection, resilience, supportability, interoperability and vendor portability.
AI Monitoring, Assurance and Operational Observability
Define telemetry, logging, audit and pre-production evaluation requirements such as acceptance thresholds for security, reliability and responsible AI requirements to monitor AI applications, model interactions, retrieval activities, agent decisions and external tool calls.
Implement or integrate monitoring capabilities to detect abnormal usage, prompt injection attempts, policy violations, unexpected tool execution, data exfiltration indicators, repeated guardrail failures and anomalous consumption patterns.
Establish security and operational metrics such as policy violation rates, guardrail interventions, attack success rates, evaluation pass rates, latency, availability, token consumption, model usage and cost.
Develop dashboards and reports that provide engineering teams and management with visibility into AI usage, security posture, operational performance and control effectiveness.
Coordinate with security operations, platform operations and application support teams to develop incident detection, escalation, containment, investigation and post-implementation review procedures for AI-related events.
Review production evidence, security telemetry and assurance findings to identify control weaknesses, recurring failure patterns and opportunities to improve AI architectures and engineering standards.
Secure AI capability development
Uplift A*STAR's proficiency in secure AI engineering and implementation. This would be achieved through hands-on workshops, labs and technical clinics covering secure AI application development, prompt and context security, RAG protection, agent tool security, AI gateway controls, threat modelling and AI security testing.
Develop a practical sec