Work
I started working part-time for the ride pooling startup CleverShuttle in 2016. Back then the
team consisted of only ten employees. In the beginning, I mainly worked on launching three new
cities. This went from writing applications for approval of the respective city government to
finding locations for the future city headquarters.
As the number of employees started growing and hence communication and organizing a bigger team
became more and more important I conceptualized the process handbook. Other tasks would include
the preparation of presentations, analyzing markets and competition and compiling documents for
support of management decisions.
Furthermore, I was responsible for the recruiting process of new business development team
members. This included creating a multi-step recruiting process, writing up case-studies for
the applicants, screening of candidates and conducting interviews.
I routinely organized the quarterly retrospective of team internal objectives and key results
and communicated with internal and external stakeholders by creating newsletters and webinars.
Projects
Machine Learning: Natural Language Processing
Detecting Right-Wing Extremist Posts
During the seminar “IT Security and Privacy” I used natural language processing techniques to train a machine learning model to detect right-wing
extremist posts on Twitter. This entailed scraping Twitter for the tweets of German politicians and training two machine learning algorithms (BERT and GRU)
on the data set.
Technology stack: Python, JupyterLab, Google Colab
Toxic Comment Classification
For the seminar “Information Systems” I delivered a tutorial lesson on the topic “BERT and beyond: Cutting edge NLP models”.
This included implementing BERT to detect toxic comments in the Kaggle Toxic Comment Classification challenge data set and writing a blog post about it.
Technology stack: Python, JupyterLab, Google Colab
Prediction of Conversion and Ratings
As part of the “Advanced Data Analysis for Management Support” lecture I worked on three different machine learning tasks. In the first one I used data
that captures user activity on an online retail website to predict the conversion of clients during a session.
The second task entailed using a gated recurrent unit (GRU) to predict book ratings given a data set of book reviews.
Finally, I used a data set of Medium articles to predict the respective number of claps each article got. I used a multi-input model consisting of
two branches to combine textual and non-textual data.
Technology stack: Python, JupyterLab, Google Colab
Machine Learning: Classification
Prediction of Returns
As part of the examination process in the lecture “Business Analytics and Data Science” I worked with data from an online retailer. I used a random forest
model to predict whether a given item will be returned by the customer. The results were evaluated using a cost matrix. Furthermore, I opened the black box
of my model and analyzed variable importance.
Technology stack: R, RStudio, RMarkdown
Data Bases
Data Bases
I implemented parts of a data bank management system. Namely a buffer, index and query manager. This was part of the course “Implementation of Databases"
at Humboldt-University Berlin during the winter semester of 2019/2020.
Technology stack: C++, SQL, CLion
