The posting
AI/ML Eng Work location- Paya Lebar
- Design and build Python-based agentic AI workflows using LangGraph, CrewAI, or equivalent, with tool integration, state management, and human approval steps.
- Build an AI agent that analyses approved sources to identify and prioritise emerging fraud risks, with supporting evidence for officer review.
- Build a second agent that checks data availability and translates validated fraud risks into candidate indicators, features, and detection rules.
- Integrate both agents with FAM 2.0 using AWS Redshift, AWS Glue, and PySpark for data preparation and feature engineering.
- Deploy modular, tested AI services and APIs using CI/CD, with automated validation and release controls for code, prompts, and workflows.
- Evaluate output quality and reproducibility, implement audit trails and prompt-injection safeguards, and monitor latency, errors, and operating costs.



