Senior Software Engineer

Build the engine that turns a pile of shipping documents into a correct customs declaration, and own it in production.

Leuven • Full-time • 3 days on-site, 2 remote

Global trade runs on paperwork. We’re replacing it with intelligence.

Customaite is the trusted AI-platform for customs automation. Customs brokers, freight forwarders and logistics leaders use our self-learning platform to clear goods faster, smarter and safer: fewer errors, lower costs, audit-ready compliance, ROI in months. With strong product-market fit and rapid growth across Europe, we’re now looking for a Senior Software Engineer who builds that power, and owns it once it’s live.

The role: builder and owner

You’ll own a slice of the platform end-to-end. One week you’re designing how a new LLM extraction step should behave when the model is wrong; the next you’re reading a Datadog trace to find out why a declaration got stuck, and fixing the root cause. You’ll work closely with product, customer success and the rest of engineering. Build it, run it, improve it; and then again.

Our stack: Java 25 and Spring Boot/Quarkus, Postgres, RabbitMQ, Elasticsearch, Redis, Auth0, Temporal. Python with pydantic-ai for the LLM and ML services. Vue 3 and TypeScript on the front-end. Helm on Azure AKS. Datadog and PagerDuty to keep us honest.

What it’s really like: a small team runs dozens of services with real customer traffic. You’ll be on-call in rotation and take your turn as release owner. Some code predates you and isn’t pretty. We run Java 25 with preview features and it occasionally bites. If you want a platform team between you and production, this isn’t it. If you want to see your decisions land with customers the same week, it is.

What you’ll do

• Own parts of the domain: Take a slice of the declaration flow end-to-end: design, build, ship, run, improve.
• Build with AI, and for it: Ship LLM-powered features (extraction, classification, agents) with evals, cost and latency as first-class concerns, and design what happens when the model is wrong.
• Keep it running: On-call in rotation, release owner in turn, Datadog as your dashboard, post-mortems that fix root causes.
• Shape the architecture: Dozens of services on Java 25; refactor deliberately, call out debt explicitly, write decisions down.
• Work agentically: Claude Code or similar is part of the daily toolkit; you raise the bar for how the team uses it.
• Raise the team: Reviews, pairing, mentoring, technical interviews.

Who you are

Must-haves

• 5+ years building and running backend systems on the JVM (Java with Spring Boot or Quarkus).
• You’ve shipped LLM-powered features to production: structured output, tool calling, evals, cost and latency control.
• Comfortable owning services in production: on-call, debugging across services, observability.
• Solid with Postgres and message-driven architectures (RabbitMQ or similar).
• You use agentic coding tools daily and can show what you build with them.
• Able to work from our Leuven office three days a week.

Nice-to-haves

• Excellent English.
• Python for ML and LLM services (we use pydantic-ai and Celery).
• Kubernetes and Azure (we run on AKS with Helm).
• Vue 3 and TypeScript on the front-end.
• Elasticsearch, Redis, Auth0.
• Experience in a scaling startup or product company.
• Dutch, and a curiosity for logistics and customs.
• A small team of highly skilled, motivated people pulling toward one goal, with no layers between you and the problem.
• Real ownership of parts of the domain, and room to grow as we scale.
• An AI-native product with real traffic, on a modern stack, where new tools are tried the week they come out.
• Hybrid: three days in Leuven, two wherever you work best.
• Competitive salary, company car and performance-based bonus plan.

What we offer

Ready to change how global trade moves?

• Send your CV and a link to something you built with an LLM in it, or a PR you’re proud of, to jobs@customaite.ai, and tell us what broke in production and what you changed.