AI Engineer

NCS (Singtel Group) - AI Engineer Singapore | Posted March 30, 2026 | Full-time

Design, build, and operate production-grade agentic and GenAI systems end to end.

Agent & Application Engineering:

• Design multi-agent systems with planning and tool-use; expose via REST/gRPC APIs
• Implement tool adapters (SQL, search, document stores, web calls) with type safety
• Build model gateway integrations (OpenAI, Azure OpenAI, Bedrock, Vertex)

Retrieval & Knowledge:

• Stand up RAG services: chunking, embeddings, indexing, hybrid/vector search
• Implement ingestion pipelines for docs, tickets, ERP/CRM data
• Optimize retrieval quality with re-rankers and A/B testing

Quality & Testing:

• Version and test prompts and graphs as code
• Build evaluation harnesses for latency, cost, accuracy, hallucination detection
• Implement drift detection and safe shutdown mechanisms

Platform & Operations:

• Deploy to Kubernetes with Helm and Argo CD
• Implement policy chains and guardrails (OPA/Gatekeeper, Presidio)
• Instrument observability: tracing, metrics, logs, cost meters

Security & Compliance:

• Manage secrets, signed images, least-privilege IAM
• Build tenant isolation and jailbreak defenses

Integration & Enterprise Workflows:

• Ship connectors for SAP/CRM/ITSM and Kafka topics
• Automate business workflows; design idempotent processors

Requirements:

• 6+ years software engineering (production services); 1+ year leading projects
• Strong in Python/TypeScript/Go/Java; comfortable in a second language
• Hands-on with containers, Kubernetes, CI/CD, IaC (Terraform), cloud platforms
• Practical LLM experience: RAG, agent apps, prompt design, safety patterns
• Data skills: schema design, pipelines, search indexes; SQL and vector DB proficiency
• Testing mindset: unit/integration tests, golden datasets for evaluations
• Security basics: secrets, policies, scanning, IAM

Preferred:

• Agent orchestration (LangGraph, Semantic Kernel) in production
• Search/retrieval tuning (hybrid search, re-ranking)
• Observability at scale (OpenTelemetry); cost/performance optimization