Hire a machine learning engineer

A machine learning engineer makes sure a model works outside the notebook. They turn experimental code into tested Python code, build the pipeline that trains and deploys the model, put the model behind an endpoint or in a batch job, and set up monitoring.

Our staffing page lists this role as a machine learning specialist, alongside roles such as data scientist, AI engineer and MLOps engineer. The right colleague depends on your model and your platform.

A data scientist works out which model answers a question and whether it performs better than a simple baseline. A machine learning engineer puts that model into production and keeps it running. An MLOps engineer sets up the shared process that all models follow: model registry, CI/CD and monitoring.

In small teams, one person often covers two of these roles. If you want that process set up as a project, read about MLOps.

Training-serving skew is the difference between how well a model performs during training and in production. A common cause is that inputs are calculated slightly differently in production than in the notebook, for example because IT has rebuilt the calculation in another language.

Google recommends using the same code in the training and production pipelines as much as possible. That is exactly what an ML engineer does: create one tested preprocessing step that runs in both pipelines.

That depends on when you need the result. If an application needs an immediate answer for each request, you put the model behind an endpoint. If a result for all records every night or every hour is enough, a batch job is easier to manage.

In Databricks, you can do both with Model Serving for models registered in Unity Catalog. Azure Machine Learning has online and batch endpoints. With an online endpoint, you can split traffic between the old and new versions, allowing you to switch safely.

Since 1 January 2025, the Belastingdienst has resumed full enforcement of the rules on false self-employment. Whether someone works as an employee in practice depends on factors including supervision, integration into the organisation and whether they operate as a business.

An ML engineer placed with you by Data Science Lab has an employment contract with us. That is different from contracting a freelancer directly. Have HR, procurement or an adviser review your own hiring policy.

If the model processes personal data, the GDPR applies. According to the Autoriteit Persoonsgegevens, a DPIA is mandatory if you use algorithmic systems involving personal data, even for a pilot. The AI Act classifies software that sorts CVs, among other applications, as high-risk; according to the European Commission, those obligations will apply from 2 December 2027.

An ML engineer sets up logging, model versioning and monitoring so you can show how a model was developed. Your legal counsel or compliance department should assess whether the rules apply to your model. Read more about AI governance.

If you do not yet have a model or a clear question, an engineer has nothing to put into production. Start with data science or a maturity assessment instead. If the problem is missing or messy data, work on the data foundation first.

If you expect to keep adding models, hiring someone permanently makes more sense. If you need a finished application rather than extra hands, have the model built as a project.