Freelance data scientist and data engineer

1/2019 – 12/2021

At de Geschillencommissie I worked on setting up a data warehouse which collects data from multiple external software providers. This data is collected using python cloud functions and published on pub sub. It's then stored and aggregated in google bigquery.
\bigskip

I also worked on a classifier model that is capable of automatically classifying new cases submitted to one of the 90 different commissions automatically. The model used was a fine-tuned dutch BERT model combined with some smart data augmentation to increase generalisation of the data set.

Google Cloud, Maschinelles Lernen, Python, SQL

12/2017 – 1/2022

At trucks.nl my initial job was setting up their search engine in elastic search and improving the free text search(including and building tokenizers and analysers) performance of their website. Later on i was responsible for a complete transition and architecture of their website and other web services to google cloud and kubernetes.

Finally I setup a data plaform that collects information from other applications they own or use and a webservice i build to collect user events from their own website. These events are then published on topics ofgoogle pub sub and stored to google bigquery using google dataflow. Using google scheduled queries I then process this and aggregated this information further and show these statistics to advertisers that publish on trucks.nl, dashboarding and other exports for further analysis.

Aggregations I worked on (using google bigquery scheduled queries) are related to extracting sessions from raw events, detecting bot traffic, calculating metrics(CTR, results, average click position) per search query, calculating activity of advertisers and advertisements, metrics per advertisement (conversion ratio, search result clicks) and how they perform.

Big Data, elasticSearch, Google Cloud, PHP, Python