Come build what runs on Monday.

AI Agent Engineer.

The work

You will build grounded agents and workflow copilots on client data, for clients in wealth operations, professional services, and cross-border SMEs. The work is real production software, not demos. Expect to own a problem end to end: scoping with the operator, grounding on their documents, shipping a reviewer-in-the-loop system, and watching the baseline move.

What you will actually do in the first 90 days

• Take a single back-office workflow from one of our clients, instrument its baseline, and ship the first working version of an agent that handles it with a reviewer.
• Design the retrieval layer over the client's private corpus (filings, SOPs, call notes). Pick the vector store, the chunking strategy, and defend both choices.
• Build the evaluation harness before the demo. No evals, no merge.
• Sit in at least one client working session per week. We do not hand off requirements through a PM.
• Write the post-mortem when something goes wrong. Publish it internally. Suggest the fix.

What we are looking for

• Four or more years shipping production backends or data systems. At least one of those years touching LLM agents, RAG, or structured extraction at real scale (not side projects).
• Strong Python. Comfort with TypeScript on the review-layer side.
• Working fluency with at least one major model provider's API and tool-calling patterns. Vendor-agnostic taste.
• You treat evals, guardrails, and observability as first-class, not afterthoughts.
• You write clearly. Code, commits, and docs.
• Comfortable working directly with non-engineer operators without a PM as a buffer.

Nice to have

• Prior work with financial operations, compliance workflows, or document-heavy back offices.
• Experience with workflow orchestration frameworks (Temporal, LangGraph, custom) and their failure modes.
• A public artifact we can read: a repo, a write-up, a post-mortem.

What you will not find here

• A layered hierarchy. We are small. You will have real scope from day one and real accountability.
• A growth-at-all-costs narrative. We grow when we have work and people, in that order.
• Work on cybersecurity or games. Out of our registered scope.

Compensation

Competitive Singapore market base, calibrated to experience. Annual bonus tied to firm outcomes, not utilisation. Meaningful equity for senior hires. Work from home two to three days per week, office the rest.