Agentic AI Engineer

Why this role is different

• A real environment, the highest standard. Our customers are Japanese financial institutions, where security, compliance and quality are non-negotiable. Experience running agents in such an environment is rare in the market.
• You build a platform, not just projects. You build the tools and components that let later agents be built faster and run more safely. What you make today is reused across many business processes and many customers.
• Always-on agents. This is not Q&A chatbot work. You build agents that run continuously, carry out multi-step business tasks on their own and call tools inside enterprise systems. This is the technology direction that will decide the industry's future.
• The whole AI lifecycle. The work spans data, fine-tuning and model optimization, through building agents, evaluation, guardrails and deploying AI agent applications to production. You are not confined to one small piece.
• Work with Japanese finance experts. You learn banking and finance directly from people who know the industry's laws and practice.
• Join from the foundation stage. You help decide the architecture, the technical standards and how the whole team will work from here on.
• Truly AI-native. Every engineer is equipped with Claude. Not knowing Japanese is no barrier to working with Japanese documents and experts.

What you will do

• Design and build AI agents, multi-agent systems and workflows for banking and finance: planning, tool use, knowledge retrieval, multi-step execution
• Develop skills, tools and workflows that can be packaged, tested and reused across different AI agent applications
• Build agent evaluation suites: golden datasets, regression tests when models or dependencies change, version comparison. Non-deterministic behavior must be measured with evidence
• Help build guardrails on a policy-as-code approach. Every model call and tool call must pass through a control layer
• Work on model pipelines: data curation, fine-tuning, distillation, quantization, serving on GPUs
• Take agents from POC to production: packaging, deploying on Kubernetes, monitoring and improving after go-live

Who we are looking for

• Has built and shipped at least one LLM application or AI agent to production, not just a demo, and dealt with the problems that appeared after going live
• Strong Python, writes clean, tested, production-quality code
• Deep understanding of how agents work: tool use, context engineering, orchestration, LLM weaknesses
• A measuring mindset: does not trust a "feeling that the agent works", and proves it with evals
• Cares about agent safety: prompt injection, data leakage, permissions for agents
• Proactive, and comfortable with the ambiguity of the foundation-building stage

Nice to have

• Project experience in banking, finance, insurance or with Japanese customers
• Experience with fine-tuning or model optimization, MLOps, GPU operations
• Experience with Kubernetes, AWS, IaC
• Good Japanese or English

Benefits

• Competitive salary
• 13th-month salary under the Labor Law
• 40-hour week, Saturdays and Sundays off
• Opportunities for business trips to Japan

Location

3rd Floor, Golden Field Building, 24 Nguyen Co Thach, Hanoi