Software Engineer, Agentic Development

Job description

WRITE THE CODE THAT WRITES THE CODE. Most engineering roles ask you to build the product. This one asks you to build the thing that builds the product. We run an award-winning platform that handles visas, work permits, business travel compliance and global expansion for companies operating in 170+ countries. The domain is genuinely complicated - immigration rules change weekly, every country is its own edge case - and our engineering team is small on purpose. We’re building our platform agentically, but agents are only as good as the harness around them: the context they're given, the tools they can reach for, the tests that catch them, the infrastructure they deploy into, the review gates that stop a confident wrong answer reaching a client. That harness is your job. WHAT YOU'LL ACTUALLY DO - Build and sharpen our agentic development setup. Context pipelines, tool definitions, sandboxes, prompt and skill libraries, eval suites, review gates. Make it fast, make it trustworthy. - Turn recurring engineering work into repeatable agent-run workflows — and be the person who says when a task shouldn't be handed to an agent at all. - Write real product code when it matters. We’ve got an eclectic stack including NextJS, Python, React Native, Java and Angular, we use Ai agents extensively, mainly Anthropic with langraph. You have to be ready to dive in. - Debug the hard things. The failures here are rarely a stack trace. They're a subtly wrong assumption three layers down, at 4am, in a country you've never worked on. - Work shoulder to shoulder with product. Sit in on customer conversations, talk to the legal and operations experts who know the domain, and turn what you hear into something buildable. We'd rather you asked a client directly than guessed. - Have opinions about architecture — and be able to argue for them without needing to win. WHAT WE'RE LOOKING FOR Non-negotiable - Around two to four years building complex web applications that real people depended on. - Strong TypeScript and Next.js. Comfortable enough in Python to be useful. - Real AI integration experience. You've shipped something LLM-powered that people used, and you can talk honestly about where it fell over. - Fluent and clear-eyed about agentic coding. You know what it's brilliant at and where it quietly costs you a day. Neither a sceptic nor a true believer. - Clear communication, written and spoken. You can explain a technical trade-off to someone in legal or sales without making them feel small. - Ownership. When something's yours, it's yours — including the bit where you tell people early that it's going sideways. - Warm, direct, curious, up for getting stuck in. Nice to have - Evals, observability or LLM-in-production tooling as a genuine interest, not a line on a CV. - Experience in a regulated or compliance-heavy domain — immigration, legal tech, fintech, healthtech. - A design eye, or at least strong feelings about interfaces. - Time in a small team where the roadmap changed under your feet and you were fine about it. - Something you built for yourself because it annoyed you that it didn't exist. WHY CENTURO Our mission is a borderless world of opportunities: making it possible for people to live and work anywhere. Global enterprises trust us to move their people across more than 170 countries, and they choose us over legacy providers because we're faster, clearer on price, and better at the complicated cases. We're human-led and tech-powered — expert lawyers alongside a platform that won Most Innovative Use of Technology in Global Mobility at the 2025 FEM EMEA awards. Every improvement you make to how we build lands on someone waiting on a visa decision. HOW HIRING WORKS 1. Intro call, 30 minutes — with our CTO. What you've built, what you want next, what this role really involves. 2. Technical conversation, 60 minutes — bring something you've built with agents, or a project you're proud of, and walk us through the decisions. No whiteboard algorithm puzzles. 3. Half a day in the office — meet the product team, have lunch, decide whether you actually want to sit here every day. We aim to get from first call to decision inside 5 working days, and we'll tell you where you stand at every stage. SOUND LIKE YOU? Send us your CV or GitHub, plus a few lines on the most interesting thing you've built with an AI agent — including what went wrong. That last part matters more to us than the polished version. We hire on the basis of what you can do and how you work with people, nothing else. If you need any adjustments to the process, tell us and we'll sort it. We're happy to sponsor eligible candidates — arranging visas is, after all, what we do.