MLOps jobs putting AI into production
Machine learning has moved from experimentation to deployment, and Dutch organisations are investing in the infrastructure needed to make AI work at scale. As adoption accelerates across fintech, insurtech, energy and transport, they need MLOps and machine learning professionals who can build, deploy and maintain production-ready AI systems. We connect experienced professionals with employers across the Netherlands, helping them find opportunities where they can work with modern technologies, solve complex engineering challenges and help shape the next generation of AI-powered products and services.
Establishing your career in MLOPs
Expert advice from Country Director, Jo Callaghan
MLOps is a production engineering discipline first, so software engineering experience with genuine ML exposure usually counts for more than academic credentials. Research backgrounds matter most only in roles close to model architecture itself. Dutch employers hire for the ability to ship and maintain a reliable system, and a recognised cloud certification, whether Azure, AWS or Google Cloud, can help a CV through the earlier, often HR-led screening in a way a research paper rarely does.
Software engineering, data engineering and data science are the strongest starting points, roughly in that order, and both university (WO) and applied sciences (HBO) graduates move into the field. Python, a cloud platform and Docker are the baseline almost every employer expects, even at junior level.
Production experience is the hard part to prove without a commercial role. A model trained only in a notebook isn't distinctive enough; deploying and monitoring something, then fixing it when it breaks, is what gets your foot in the door. Employers want judgement above all, telling genuine drift from noise, and a well-documented real incident tends to land better in interview than a list of tools, since hiring managers read a clear postmortem as a sign of engineering maturity.
Python is close to universal here, but what separates a strong MLOps or machine learning engineer from a data scientist who can code is production engineering discipline: the ability to design a system that keeps working when the data underneath it changes. Here's the skills which we consistently see across the briefs we work:
- Python: the standard language across nearly every part of the stack, from model training to pipeline orchestration.
- Docker and containerisation: packaging models and their dependencies so they run the same way everywhere, from a laptop to production.
- Cloud computing: Amazon Web Services remains the most requested platform, though Azure and Google Cloud both come up regularly – a recognised cloud certification carries real weight at the CV-screening stage here.
- MLOps platforms and tooling: MLflow, Kubeflow and similar tools for tracking experiments, versioning models and managing the deployment pipeline.
- Model evaluation and monitoring: measuring whether a model is still performing once it's live, not just when it was tested.
- System design: architecture that handles failure gracefully, scales with demand, and doesn't need rebuilding every time a new model type gets added.
- Software engineering fundamentals: version control, testing and code review discipline, applied the same way a software engineering team would apply them.
- Front-end and back-end awareness: useful, sometimes essential, for platform-facing roles building the tools other teams use to deploy and monitor their own models.
- Model optimisation knowledge: increasingly valuable for inference speed and edge deployment, a growth area given the Netherlands' high-tech and hardware ecosystem, though still a specialist rather than universal requirement.
- Governance and regulatory awareness: the AVG and the EU AI Act's obligations for deployed, high-risk systems, since MLOps engineers are often the ones operating the systems the regulation applies to.
- Decision-making under uncertainty: knowing when a model's performance drop is genuine drift worth acting on and when it's noise, a judgement call a dashboard can’t make for you.
Generative AI has added a new layer to this. Engineers who understand how to deploy, evaluate and monitor large language models, not just call an API and hope, are increasingly valued, particularly as the EU AI Act's obligations for general-purpose and high-risk systems phase in.
Dutch MLOps and machine learning pay builds with seniority and depth in production: from the low €40,000s at junior level to roughly €65,000 for an established engineer, €85,000 or more at senior level and well over €100,000 in leadership. Engineers who can take a model from research through to reliable production tend to sit at the upper end, and freelancers typically quote around €75 to €120 per hour depending on the engagement.
| Seniority Level | Typical Permanent Salary |
| Junior ML / MLOps Engineer (0-2 years) | €42,000 – €55,000 |
| ML / MLOps Engineer / Mid-level (2-5 years) | €55,000 – €78,000 |
| Senior ML / MLOps Engineer | €80,000 – €110,000 |
| Lead / Principal ML Engineer | €100,000 – €130,000 |
| Head of MLOps / Head of Machine Learning | €105,000 – €145,000+ |
Amsterdam has the largest concentration of machine learning and MLOps employers and the highest advertised ceiling, while Rotterdam, Utrecht and Eindhoven add logistics, energy and high-tech work. With hybrid working standard, choose the production problem and sector you want first and the city second. Current market data:
| Seniority Level | Typical Permanent Salary |
| Amsterdam | €40,000 – €90,000 |
| Utrecht region | €44,000 – €78,000 |
| Rotterdam | €38,000 – €75,000 |
| Eindhoven | €34,000 – €57,000 |
| Netherlands Overall | €37,000 – €80,000 |
In practice, less than the job titles suggest. A Machine Learning Engineer typically owns the model itself, building, training and evaluating it, often through to a first production deployment. An MLOps Engineer typically owns everything around it once it's live: the deployment pipeline, infrastructure, monitoring, automated retraining and overall platform reliability, covering CI/CD, drift monitoring, infrastructure across multiple models and teams, and incident response when something breaks.
In smaller teams and Amsterdam scale-ups the two roles usually blur into one person, which is why job specs use the titles interchangeably. At larger, more mature organisations, such as banks and insurers, the split tends to be genuine, sometimes formally so, with the title tied to a defined job grade rather than just a description. Scope and seniority drive pay far more than which title is on the spec, so our approach starts with what you do day to day, not what's titled on your CV.
MLOPs & Machine Learning Career Progression
Expert advice from CEO, Jonathan Ellerbeck
Progression in this specialism tends to split along one of two tracks early in careers: staying close to the infrastructure and platform, or moving towards the modelling and research side of the work.
- Junior to mid-level: taking ownership of a full model lifecycle, from training through deployment to monitoring, rather than just one stage of it, usually over two to three years.
- Mid to senior: the point where system design starts to matter more than model accuracy. You're now responsible for what happens when something breaks at 2am, not just whether the model scored well in testing.
- Senior to lead or principal: setting the platform's technical direction and making the architecture decisions everyone else has to live with, while helping less experienced engineers get through their first production incident without panicking.
- Technical track: Principal ML Engineer or ML Architect, for those who want to keep building rather than manage a team.
- Management track: Head of MLOps or Head of Machine Learning, for those drawn to leadership and roadmap ownership rather than staying hands-on with the code.
Mentorship matters more here than the job title suggests. Production ML failures tend to be expensive and public, and engineers who've had a senior mentor talk them through their first serious incident consistently progress faster than the ones left to work it out alone.
If anything, AI is having a positive impact on MLOps and machine learning engineering work - every model or LLM a business puts into production still needs someone to deploy it, monitor it and keep it working when things drift. That need has only grown as generative AI has moved from a demo to something companies run day-to-day. The tools have changed: platform engineers now spend real time on LLM-specific infrastructure, evaluation harnesses, retrieval pipelines, inference cost management, that barely existed two years ago, but the underlying discipline, keeping a system reliable in production, is more in demand than it has ever been.
Regulation reinforces this. As the EU AI Act's obligations for high-risk and general-purpose systems phase in, and the AVG continues to govern personal data, the people operating those systems, monitoring them, documenting their behaviour and proving they still perform as expected, are the ones compliance increasingly runs through. For MLOps engineers, AI isn't competition. It's the reason the job exists, and the reason it keeps growing.
Both. The freelance market for MLOps is smaller than for some data specialisms, but it's a specialism where scarcity pushes rates towards the upper end of Dutch IT freelancing, and a production-capable MLOps freelancer rarely stays unplaced for long. Freelancers usually quote hourly, and how an engagement is structured under the Wet DBA matters as much as the rate.
Remote working suits most of this role well: designing pipelines, managing model deployment and monitoring dashboards against a cloud environment works from anywhere, and hybrid working is common across Dutch tech. The exception is edge and embedded deployment, which matters around the Netherlands' high-tech and hardware ecosystem: getting a model running reliably on physical hardware benefits from being on site for at least the initial bring-up and integration testing, even when the rest of the pipeline is built and monitored remotely.
Demand spans a genuinely wide range of Dutch industries, though the sophistication of the work varies a lot by how far along a company's AI strategy is:
- Fintech and financial services: fraud detection, trading infrastructure and credit risk models running at production scale, centred on Amsterdam, with governance and monitoring carrying weight.
- Insurtech and insurance: model deployment for pricing and underwriting, running continuously against live data under strict explainability expectations.
- Energy and greentech: forecasting and optimisation models for an energy system that is becoming smarter and more data-driven.
- Transport and logistics: demand forecasting and operational optimisation across port and logistics networks.
- High-tech and hardware: edge and embedded inference for industrial and equipment systems, notably around Eindhoven.
- Technology, e-commerce and scale-ups: embedding machine learning and generative AI directly into products, from recommendation to fraud and abuse detection.
- Public sector: operational and policy modelling roles, usually needing to satisfy the AVG and the EU AI Act before you start.
- Consultancies: MLOps and machine learning delivery across a rotating client list, often building a platform from scratch.
The strongest roles right now sit with companies that have already proven a model works in a lab and are hiring specifically to get it running reliably in production, not businesses still deciding whether AI is worth the investment. You can usually tell from the job spec alone: a named platform problem with budget behind it, versus a hope that the right hire will figure out the AI strategy for them.
