Head Office
Other
Highlight
The AI Engineer builds production agents end-to-end on an AI-native retail decisioning platform - prompt design, tool definitions, multi-step workflows on the agent runtime (LangGraph, CrewAI, or chosen framework), evaluation harnesses (golden sets, regression gates, multi-step replay), human-in-the-loop gate integration, and per-agent cost optimisation. The role consumes platform-provided LLM and vector services; it does not rebuild that platform.
Remote candidates outside of Thailand are welcome to apply.
Responsibilities
Build agents on the platform's agent runtime - prompt design, tool definitions, multi-step workflows, error handling - and ship them with eval harness, human-in-the-loop gate config, observability instrumentation, cost meter, and runbook.
Co-design agent specs with Tech Lead Applications and Suite Product Owners; partner with ML Engineers on classical ML model integration into agents.
Author golden sets per agent - domain-specific test cases capturing must-pass behaviours; build regression gates in CI so no agent ships without eval-pass.
Implement multi-step conversation replay for agents with stateful interactions; use LLM-as-judge patterns where appropriate; instrument human feedback collection.
Configure HITL gates per agent and per agent plan; implement gate-progression evidence collection (Shadow data, accuracy metrics, override frequency).
Own per-agent cost meter - tokens, vector queries, model inference; report monthly; tune model routing and implement caching strategies where appropriate.
Consume the enterprise LLM Gateway via standard SDK; partner with platform AI engineering on embedding model selection and retrieval relevance tuning.
Mentor seed-programme engineers and contribute to the agent-engineering playbook.
Qualifications
Bachelor's or Master's degree in Computer Science, AI / ML, or a related discipline.
5+ years software engineering with 2+ years shipping LLM-based or agentic systems to production.
Production agent or multi-step LLM workflow experience - LangGraph, CrewAI, AutoGen, DSPy, or custom.
Strong Python; comfortable with async, observability, testing.
Hands-on with at least one major LLM provider (Azure OpenAI, Anthropic, Bedrock, Vertex).
Eval-driven LLM development - golden sets, LLM-as-judge, regression gates, multi-step replay.
HITL gate / agent governance - has shipped agents with explicit gates, not autonomous-by-default.
Prompt injection / data leakage / PII handling - designs and tests defences.
Open-source contributions to agent frameworks (LangChain / LangGraph / DSPy).
Multi-agent system at scale in production; retail / commerce / fintech agentic workflows (supplier onboarding, contract intelligence, comparable).
Causal inference exposure (DoWhy / EconML); Thai-language NLP (PyThaiNLP, WangchanBERTa, SEA-LION, Typhoon).
Vendor certifications such as Databricks Generative AI Engineer or Azure AI Engineer Associate.