KPMG IPay

In Short

Refactoring of an older solution that focuses on Big Data Analytics. The original solution involved mostly around the apache-stack (Stream, Mongo, Spark), whereas the new solution replaces these components where possible by Azure-based components.

IPay

The IPay project focused on predicting the value of un-used gift cards. Getting a giftcard for your birthday can be great, but sometimes we forget about them and end up throwing them away after they are no longer valid.From an accountancy perspective, it might be interesting to have an idea about the amount that is laying around in people’s houses that is not going to be used anymore. This way, it is possible to lower the reserves needed to cover the potential future usage of those cards. And it appears that this is also quite predictable, as long as you know how to properly model it.

Technical Details

Without going too much in details, the original creation of the data processing platform was mostly through Hadoop, Spark and MongoDB. Back at the day, you would go through a lot of pain to set these up individually, including observability, change management, security, etc, etc. Luckily, it is becoming more and more common to replace these components by more managed versions through Cloud-offered services. That was exactly the idea behind the project. Migrate the big data processing platform to Azure. This had quite some caveats though, most importantly the use of the C++-based Data Analysis Framework developed by CERN called ROOT, which still exists and has a lot of applications within science.

My Responsibilities

I was involved with various data engineering aspects, mostly focusing on Azure, Python, Spark and containerizing analyses.