Machine Learning in Production

Making Machine Learning add value

There is no value to invest in data and ML unless they are used in production. To make this happen, data needs to be readily available and stored efficientlyMoreover, models need to be updated continuously to keep them from degrading, checked that they behave as expectedtheir deployment orchestrated, and fail graciously in case of problems. Even more, often models need to scale to efficiently serve many customers and work with huge amounts of data. 


According to O’Reilly we have entered a new era of data science where machine learning systems have become an accepted part of delivering value to customers.

And already in 2015 Google highlighted in a widely-cited paper that such systems have many components besides the models they employ, which touch upon data, compute resources, efficiency, way of working and much more.

The hard work of building ML-powered data products

With the increased use of data and ML in organizations, and with that increased complexity and responsibility, it is no longer possible to ask all this to be part of the skill set of all your professionals. Instead, you need people who focus on certain parts of the setup (e.g. data scientists, data engineers, or data analysts). And the professional who focuses on productionizing  models and data requirements is called machine learning engineer. 


Engineering the ML way of working

Machine learning engineers (MLE’s) have a deep knowledge of both Machine Learning and Software Engineering. They understand that models are living artifacts, so they engineer systems that automatically retrain models as the data changes and track models over time. This includes automatically flagging data quality issues which may affect not only the accuracy of the models, but also their fairness.  

MLE’s also play an important role in defining a way of working for model development, ensuring that new features can easily be added to models and that new model versions can be deployed quickly and reliably. Taken together, these practices are usually referred to as MLOps. 


Daniël Heres is a Machine Learning Engineer at GoDataDriven. He has experience building data-driven solutions at companies such as, HEMA, KPN and Ahold. As a Machine Learning Engineer, he bridges the gap between data engineers and data scientists. Daniël holds an MSc in Computer Science and enjoys creating solutions that bring value out of data using scalable and maintainable software.

Get in touch with the experts

Let's discuss the next step

Are you interested in how we can help your organization? Contact our Sales Manager Tim Waijers if you want to know more. He’ll be happy to help you!