An agentic system is only as good as the surface a human meets it through. Somebody has to approve the exception, understand why the agent chose what it chose, and correct it when it is wrong. Build that surface badly and a technically sound system gets quietly abandoned.
This role owns those interfaces across our engagements. It is remote within Asia, with deliberate overlap into European hours so you are working with the forward deployed team rather than behind them.
What you will do
- Build review queues, approval flows, and human-in-the-loop interfaces for deployed agentic systems
- Design trace and explanation views that make an agent's reasoning legible to a non-technical operator
- Own the front end and the API layer behind it, in React and TypeScript against Python services
- Work directly with client end users, watch them use what you built, and fix what you see rather than what they report
- Build the internal tooling our own engineers use to run engagements
What we look for
- Strong React and TypeScript, with real care for how an interface behaves under latency and failure
- Backend competence in Python, enough to own an endpoint rather than wait for one
- Product instincts. You have opinions about empty states, error copy, and what a streaming response should feel like
- Comfortable working across time zones with several hours of European overlap
- You have watched a real user struggle with something you built and changed your mind because of it
Nice to have
- Design skills, or a track record of shipping without a designer
- Experience with streaming interfaces and token-level UX
- Accessibility work on data-dense internal tooling
Your first six months
- Month 1: ship a review interface into a live engagement
- Month 3: own the product surface across two or more engagements
- Month 6: our interface patterns are yours, and new engagements start from them