13 June 2019 ... min. read Listen

Machine Learning as a Service

ING’s Marcin Pakulnicki, Pierre Venter and Effi Bennekers reveal how ING is building a Machine Learning Model Serving platform to enable IT teams to tap into the power of machine learning.

Effi is an engineer on the Machine Learning Model Serving platform. He will explain how IT teams within ING are able to utilise the platform for their specific projects. Marcin and Pierre had just such a project, and they will be presenting their use case at Codemotion.

Marcin and Pierre are both software engineers for ING online banking and the mobile banking app. Thousands of user reviews, many of them IT-related, are submitted through these two channels. All the reviews have to be categorised and assigned to the right teams who work on resolving them. Until recently, the reviews were categorised manually. Machine learning (ML) has enabled a dramatic reduction in the amount of time it takes between receiving the feedback and updating the feature.

Substantially shorter lead times

Marcin: “Before we used machine learning, a couple of people had to plough through all those thousands of reviews to group them together and pass them on to the right team. It was not only a long-winded and inaccurate process, but it also took a really long time before the issue was resolved.” Pierre continues: “The ‘new’ feature was sometimes outdated before it even went into production. It was clear that the lead times needed to be substantially shorter.”

Thanks to machine learning, the reviews are now algorithmically classified in real time using ML libraries and are immediately forwarded to the relevant team. Pierre and Marcin thought about what they needed, and Effi and his team made sure that this platform would enable them to deliver it. Pierre: “We gathered data and applied the algorithm for machine learning. We used that to train the first model and then rolled out a proof of concept.”


Marcin: “Our specific problem is just one of the many issues that can be solved using machine learning. We were keen for our use case to encourage teams elsewhere within the bank to think how the ML platform could offer them solutions. And at Codemotion we hope to inspire other organisations to do the same!”

Rapid classification and follow-up of customer feedback

Live on stage

The trio is often speaking at conferences. In their presentation, they demonstrate how you can make machine learning scalable within your own organisation. Marcin explains the background to the use case, and then Pierre will create and train the model live on stage before demonstrating how it classifies unlabelled data. The audience will be encouraged to submit reviews on the spot. Effi: “Then it’s over to me. I talk about the platform itself and how you can use a generic solution for your own specific model and data. Our key message to IT teams is: make use of what’s already available! Don’t reinvent the wheel or buy yet another more extensive software package. Keep it simple and make sure that your platform serves the needs of your organisation’s engineers and data scientists.”

Open source

Whenever present they hope for lots of interaction with the audience. “Our case is relevant for any company that has customers – so, for pretty much everyone. Our main aim during the conference is to share knowledge and gain new knowledge in return. That will help us all to advance the machine learning field together. In the longer term we also want to make the platform open source so that any company can benefit from it.”

Pierre Venter

Pierre, who is from South Africa, has been working at ING for the past three years. He didn’t make the switch to IT until he was in his thirties. Before joining ING, he worked at Siemens as a systems engineer. He feels right at home at ING, an environment in which innovation, tech and customer focus go hand in hand. Pierre is a passionate software engineer who has a Chapter Lead role at ING.

Marcin Pakulnicki

Marcin taught himself to code and originally started out as a Flash developer. He is crazy about new technology and his passion lies in making technical innovations scalable. Marcin joined ING two years ago. “ING is a fantastic company for engineers; in terms of IT, we are way ahead of other organisations.” He is keen to tell the world about this, which is why he is speaking at ever-more external events and conferences. Marcin is a big fan of science fiction films and Eastern European sci-fi authors such as Stanislaw Lem.

Effi Bennekers

Effi started coding when he was just six, helped out with IT-related matters at his father’s company from the age of 15 and started his professional career in multimedia solutions and building websites. He has been working at ING since 2013, since the summer of 2018 as an engineer on the Model Serving platform. Effi believes it is very important to share knowledge and it also happens to be one of his favourite activities. He launched two communities within ING: the Back-end Guild and the Applied Machine Learning Community. Outside of work, he enjoys brewing his own beer.

Back to top