M02 - ML/AI Engineer (FAM)

Role Overview The ML/AI Engineer is embedded within the data analytics and AI division, working as part of the division's fraud analytics team. The role supports fraud analytics platform, which detects misuse of public training subsidies, by designing and building two agentic AI systems. The first surfaces emerging or blind-spot fraud risks. The second converts validated risks into actionable detection logic. Human validation, evaluation and responsible-AI guardrails are embedded throughout. The engineer is expected to deliver working, evaluated MVPs and to build up DSAD's in-house capability to operate and maintain these systems independently. Key Responsibilities • Agentic AI Architecture: Design the end-to-end architecture for fraud analytics agentic AI solutions, including agent orchestration, workflow design, state management, tool integration, human approval points and system interactions, using suitable frameworks such as LangGraph or CrewAI. Architecture decisions are subject to consultation with the platform team. • Emerging-Risk Agent Development: Develop AI agents that analyse approved information sources to identify, generate and prioritise evidence-linked emerging fraud and abuse risks for officer review. • Detection Engineering Agent Development: Develop AI agents that assess data feasibility, propose candidate risk indicators and detection features, and support the translation of validated risks into actionable analytics and detection logic. • Data & Analytics Integration: Integrate agent workflows with fraud analytics platform data and analytics components, including Redshift, Glue and PySpark. Build and maintain data preparation and feature-engineering processes required to support AI and ML workloads. • Production Engineering: Convert prototypes into modular, tested, maintainable and deployable services, workflows, APIs or agentic applications, with clear interfaces, version control and operational ownership. • CI/CD & Automation: Build automated workflows for testing, validation and promotion of code, prompts, workflows and AI components across environments, incorporating quality, security and human approval gates where required. • Evaluation & AI Guardrails: Implement evaluation frameworks and governance controls covering output quality, reproducibility, traceability, access control, exception handling, human approval gates and safeguards against untrusted inputs such as prompt injection. • Monitoring & Reliability: Implement observability, dashboards and alerts covering system health, latency, throughput, errors, resource utilisation, data quality, workflow performance and operating costs. Support incident management, troubleshooting and root-cause analysis. • Domain Collaboration: Work closely with data scientists, business users and analysts to capture domain expertise, refine agent workflows and ensure solutions address operational fraud detection needs. • Documentation & Knowledge Transfer: Produce maintainable code, architecture decision records, evaluation evidence, operational runbooks and technical documentation, and conduct knowledge transfer to support long-term sustainability of the solution. Education & Experience • Education: Degree in Computer Science, Artificial Intelligence, Data Science, Software Engineering, Information Systems or a related technical discipline. • Experience: 3-5 years of hands-on experience building and deploying ML/AI applications. This must include at least one LLM-enabled system delivered beyond prototype into operational use. Technical Competencies • Strong Python software engineering capability, including API development, testing, version control and modular system design. • Hands-on experience building and deploying ML/AI applications, including at least one LLM-enabled or agentic AI solution beyond proof-of-concept stage. • Experience with LLM applications, retrieval, tool use, workflow orchestration or multi-agent frameworks such as LangGraph, CrewAI or equivalent. • Experience integrating AI/ML solutions with enterprise data platforms, databases, APIs or cloud services. • Ability to design and implement systematic evaluation frameworks for AI systems, including testing, benchmarking, observability and reproducibility. • Working knowledge of AI governance, security and Responsible AI practices, including human-in-the-loop controls, access management, auditability and safeguards against misuse. • Strong problem-solving and analytical skills, with the ability to translate business requirements into scalable AI-enabled solutions. Preferred Qualifications • Experience with AWS services such as Bedrock, Redshift, Glue, S3, Lambda or related cloud-native services. • Experience with PySpark, Databricks or large-scale data processing and feature engineering pipelines. • Experience with CI/CD, containerisation, MLOps or model deployment tooling (e.g. Docker, MLflow, GitHub Actions, Jenkins). • Experience implementing monitoring, observability and operational support processes for ML/AI applications. • Experience delivering fraud detection, risk, compliance, investigation or decision- support solutions. • Experience with knowledge retrieval, RAG, graph-based workflows, prompt management, model evaluation tooling or AI observability platforms. • Familiarity with Singapore Government or public-sector technology environments, including data governance, security and compliance requirements. • Experience working with cross-functional teams comprising data scientists, engineers, analysts and business stakeholders. Soft Skills • Strong engineering judgement and structured experimentation, with the ability to distinguish a robust MVP from a demonstration. • Ability to explain architecture, evaluation evidence, risks and controls to technical, governance and management audiences. • Collaborative working style with fraud-domain officers. The engineer should design systems that capture their expertise while leaving decisions and accountability with them. • Ability to integrate quickly into an existing team and work effectively with limited supervision. • Comfort working across research, engineering, integration, documentation and stakeholder engagement. #J-18808-Ljbffr