AI AI-Native Engineer Hoofddorp · hybrid

Devleaps; Hoofddorp; Hybrid; EUR 85,000 - 140,000; Full time

Job summary

Devleaps, an AI software partner, seeks an AI-native engineer who builds with agents daily and cares about evals, guardrails, cost and adoption as much as the model itself. You will design AI harnesses, create scalable AI factories, and deliver applied AI solutions tailored to client contexts and production environments.

You’ll collaborate with client engineers, run hands-on sessions, and help organisations move beyond pilots toward production-ready AI workflows and governance.

Qualifications

• 4–8 years of software engineering experience
• 1–2 years building with LLMs, agents or generative AI systems
• Fluent Dutch and professional English

Responsibilities

• Build the building blocks for AI: reusable skills, blueprints, prompts, tools and MCP servers
• Design and build AI harnesses with the organisation's standards and policies
• Help build AI factories across production lines, with gates from intake to release
• Build applied AI solutions: agents and RAG systems that fit client needs
• Build the gate, not only the generation: evals, guardrails and cost attribution

Skills

Software engineering

LLMs / AI systems

Python / TypeScript

CI/CD

Cloud platforms

Dutch language

Communication

Tools

LangGraph

AI SDKs

OpenAI / Azure OpenAI

Job description

You help organisations get the most out of AI by engineering it into how they actually work, instead of leaving it in pilots and on laptops.

At Devleaps we help clients move from "everyone has a coding agent and we ran a chatbot pilot" to an organisation where AI is part of the way software gets built and the way the business runs. Sometimes that means building an agent that answers a real business question on top of the client's own data. Sometimes it means building the AI Software Factory around their engineering teams, with a harness that feeds agents the organisation's standards and gates that decide what is allowed to reach production. Most of the time it means both, and in every case it means leaving behind building blocks the client's own people can reuse long after we are gone.

We are looking for an AI-Native Engineer who works that way themselves. You build with agents every day, you have opinions about what makes a harness good, and you care just as much about evals, guardrails, cost and adoption as about the model underneath.

WHAT YOU'LL DO

• Build the building blocks for AI: reusable skills, blueprints, prompts, tools and MCP servers, retrieval components and guardrail blocks that teams can pick up and trust, rather than every squad reinventing its own version on a laptop.
• Design and build AI harnesses: the environment an agent works in, with the organisation's real context, its standards and policies, scoped permissions, the right tools and a memory that carries over from one run to the next, so what the agent produces matches the real world instead of a generic sandbox.
• Help build AI factories: implement the AI Software Factory at clients across its whole production line, from typed intake and planning through implementation, policy and eval gates at review, and release into existing CI, all the way to monitoring where every run is traced and its verdicts feed the next one.
• Build applied AI solutions: agents, agentic workflows and RAG systems that solve an actual business problem, built on whatever AI foundation the client already has, whether that is Azure AI Foundry, Bedrock, Gemini Enterprise or orq.ai.
• Build the gate, not only the generation: evals, policy checks, input and output guardrails, audit trails and cost attribution per team, repo or ticket, so the client can see what was checked and what it cost.
• Measure it honestly: baseline where a client stands and track autonomy level, flow efficiency, review time, escaped defects and rework, because throughput on its own tells you nothing.
• Make people AI-native: pair with client engineers, run hands‑on sessions and hackathons, and take the champions and the teams that stalled along together, so the way of working sticks after the engagement ends.
• Stay grounded: know when a good prompt beats fine‑tuning, when a deterministic check beats another agent, and when the honest advice is not to use an LLM at all.

YOU ARE

• AI-native yourself: agents are part of how you work every day, you have built your own skills and workflows, and you can explain from experience where they help and where they fall over.
• An engineer first: you care about system design, reliability and maintainability, and you see AI as the domain rather than a reason to skip the fundamentals.
• Foundation-agnostic: you choose the model, framework and platform that fit the problem and the client, without being religiously attached to any one vendor.
• Someone who thinks in systems: you look past the single agent to the harness around it, the loop that makes it learn, and the controls that let an organisation trust it.
• A bridge‑builder and a teacher: you are as comfortable in front of a CTO explaining why review is the real bottleneck as you are pairing with a developer on their first agent workflow.
• Curious and honest: the field moves monthly, you learn fast, you know what you know, and you never oversell what the technology can do.

YOU'LL BRING

• 4–8 years of software engineering experience, with at least 1–2 years building seriously with LLMs, agents or generative AI systems, ideally something that runs in production.
• Daily, hands‑on experience with coding agents such as Claude Code, Cursor, Copilot or Codex, including writing your own skills, commands, rules or agent configurations.
• Strong Python or TypeScript, and experience with LLM APIs and agent frameworks (Anthropic, OpenAI, Azure OpenAI or Bedrock; LangGraph, the OpenAI or Claude Agent SDKs, or similar).
• Experience with tool use and MCP, multi‑agent orchestration, memory and context engineering, and human‑in‑the‑loop patterns.
• Experience building and evaluating RAG: chunking, embeddings, re‑ranking, retrieval quality and hallucination mitigation.
• Hands‑on work with evals, guardrails and LLM observability (tracing, cost and quality metrics), with tools such as Langfuse, LangSmith, Braintrust, orq.ai or similar.
• Comfort with CI/CD, containers and at least one cloud (Azure or AWS preferred), because gates and harnesses have to plug into the SDLC a client already runs.
• Experience with developer platforms, golden paths or DevEx tooling is a strong plus, as is having run a workshop, a hackathon or an adoption programme.
• Fluent Dutch, professional English.