Free with a WorkJio account: we compare your profile with this job and show what you already meet.
Share with a friend
Start
Not stated
Work mode
On-site
Experience
Experienced
About the role
You will join AssuranceAI, the AI Programme's function focused on AI behaviour and AI risk, and own an assigned product area covering evaluation, runtime safeguards, assurance reporting and AI manipulation detection. Reporting within a team led by a lead product manager, you will work with a high degree of independence in your scope, getting into the details with engineers and policy teams. This suits someone with product experience who is curious about AI risk and online safety and keen to go deep quickly.
What you'll do
Execute the product roadmap for AI evaluation, covering groundedness, citation quality, safety and adversarial testing, and regression evaluation
Product Manager, AssuranceAI (AI Safety, Evaluation & Governance), AI Programme job at GovTech Singapore, Singapore | WorkJio
Work with engineering to build and maintain reusable evaluation sets for product teams across the AI Programme
Use AI observability data to support groundedness evaluation, unsafe output analysis and AI risk reporting
Contribute to runtime safeguarding capabilities such as unsafe prompt and output detection, sensitive-content controls, prompt-injection indicators and policy-based AI controls
Work with engineering to embed safeguards into development workflows under the assurance-by-design approach
Support real-time monitoring for model drift, unsafe outputs and prompt-injection attempts, flagging issues and working with engineering to resolve them
Maintain and improve AI assurance reporting, including evaluation history, risk and remediation records and assurance status
Track and report on KPIs such as evaluation coverage, guardrail efficacy, mean time to detect vulnerabilities and policy coverage
Execute the product roadmap for a whole-of-government platform that benchmarks and integrates AI detection solutions across image, video and audio
Support continuous evaluation as the deepfake threat landscape evolves, gathering performance data and contributing to decisions on which detection capabilities to onboard, refresh or retire
Work with capability developers and partner agencies to keep the platform operationally effective as a shared government service
Participate in the Community of Practice across government, sharing benchmarking lessons, datasets and evaluation insights
Translate policy direction and requirements into product specifications, evaluation criteria and platform features
Work with the AI Practice to keep teams current on the AI risk and vulnerability landscape, and contribute to playbooks and guidance materials
Good to know
Flexible work arrangements are available, subject to the job role
Learning and development opportunities are provided year round
Requirements
Has owned products or feature areas end-to-end and shipped things that work for users
Genuinely interested in AI risk, evaluation, safeguards or online safety and willing to learn quickly
Comfortable working with evaluation metrics, benchmarking results and usage data to make decisions
Tests assumptions before committing to development and uses evidence to navigate trade-offs
Able to read policy direction and translate it into requirements, features and platform decisions
Works effectively with engineers, agency users, capability developers and policy teams, keeping them aligned
Benefits
Leave benefits to meet work-life needs
Employee wellness programmes
Flexible work arrangements, subject to the job role
Continuous learning and development opportunities year round
Apply on the company website
You will leave WorkJio. Never pay a fee or share your Singpass password, OTP or bank log-ins to apply.