AI Solution Engineer

• Understand before you build. Sit with the people doing the work, learn and map how it really happens, figure out where an agent genuinely helps and where it is the wrong tool.
• Build and ship in their stack. The agents, the connections into their data and systems, and a usable interface when the job needs one. You write the code, and it runs in production, not in a demo.
• Draw the line. Where the agent acts alone, where a person decides. Shape the handoffs so people trust the system rather than fight it.
• Ship to adoption, then make yourself unnecessary. The client’s team can run, trust and extend what you built, so it keeps working long after you have moved on. What survives without you is the real measure of success.
• Carry it forward. Same problems, different disguises. Turn repeating parts into building blocks the team reuses. Each engagement should starts further ahead than the last.
• Own the playbook. We’re early in this, so there isn’t one yet. What you learn in the field becomes how Dataroots offers and delivers agentic AI: the building blocks, the patterns, the arguments we take to clients., * Breadth (front-end, cloud setup, integration glue, etc.)
• Familiarity with agentic toolkit (MCP, retrieval, knowledge graphs)
• Background close to customers or between product and users (consulting, solutions engineering, early-stage product work)

The Offer

• An attractive salary with extralegal benefits, including:
• Mobility budget or a company car with fuel/charging card
• Hospitalization and group insurance
• High-end laptop
• Smartphone with subscription
• Substantial amount of holidays
• Meal vouchers
• …
• Diverse and welcoming work environment, where you’ll collaborate & unwind with colleagues from different cultures and disciplines. The organization of both fun & professional events and initiatives is actively encouraged and supported.
• A training budget for individual and team learning opportunities.
• Tons of team-building events and sports initiatives to stay connected and unwind.

A capable engineer to ship things people depend on. We hire for judgment and drive rather than a complete checklist, and we look for people who can create momentum from a thin brief. * Engineering principles. ~3 years of building software/data systems. You have shipped code to real users, prefer simple solutions and can own a system E2E. (Python is a plus)

• AI mindset. You take LLM tools and systems past the demo. You know where they work, where they are limited, and how you’d prove it in production: evaluation, observability, the unglamorous parts that keep an agent honest. (If most of that mileage came from your own projects, side builds and weekend experiments, that counts)
• Comfort in chaos. You do well when the brief is thin and the ground is half-built. You bring structure and momentum instead of waiting for someone to hand you clarity.
• Why, not what. You think about why a client wants something, not just what they asked for. You’re willing to challenge, think out of the box, and propose a better path.
• Build to last. Build tools to survive the handover, with the customer instead of around them. You communicate clearly enough to be trusted by both technical and non-technical stakeholders.
• Signal ownership: founder, first engineer, technical lead, or the person everyone came to when something had to work by Friday.
• Fluent English, plus Dutch and/or French.

Shift and schedule

Weekend availability

A client decided agentic AI is the way forward. Budget approved, sponsor enthusiastic, slides finished. Nothing runs yet.

That’s where you land. You turn the intent into something that works inside the client’s own world: their data, their systems, their identity setup, their people, their constraints.

You sit in Strategy & Analytics, between the business and the deep technical teams. Half of this job is engineering; the other half is spending real time with the people who will use the system, learning how the work actually happens today, and deciding where an agent genuinely helps and where it is the wrong tool.