2019 - today: exercism

Customer: exercism.io

Tasks:

exercism is a platform for learning programming languages. It is based on student-mentor interaction, where each mentor is a voluntary. I have been mentoring Python and Ruby code so far and helped 60+ students.

2019 - today: Data Engineering & Infrastructure for Mechanical Engineering

Customer: grandcentrix GmbH

Tasks:

In this project I work as both data engineer and cloud-infrastructure developer to implement the data science platform for a large German mechanical engineering company. The infrastructure is entirely written in Terraform and hosted on Azure, using i.a. Databricks and Kubernetes clusters.

2019 - today: Python Library for Business Intelligence

Customer: grandcentrix GmbH

Tasks:

In this project I designed and implemented a Python library for integrating different data sources via RESTful APIs using asyncio and the aiohttp client. The design relies on Python?s object orientation principles (e.g. duck typing) and aims for both readability and stability.

2018 - 2019: Geospatial backend for dashboard vizualisations

Customer: grandcentrix GmbH

Tasks:

In this project I implemented a dashboard application in Django for querying and displaying geospatial, sensor, and KPI data for connected vehicles. The solution is based on the GeoDjango and pandas frameworks as well as the PostGIS plugin for PostgreSQL and hosted on Microsoft Azure. Given that this was my first large data project, the biggest challenge was database and query optimisation.

2018 - 2018: A Decompositional Approach to Stance Detection on Twitter

Customer: Bielefeld University, AG Semantic Computing

Tasks:

In my M.Sc. thesis I conducted original research on (political) stance detection with a focus on the #MeToo movement. I used i.a. NLTK, pandas, and scikit-learn to train classifiers on self-annotated Twitter datasets.

2017 - 2017: Analyzing Political Stance

Customer: Bielefeld University, AG Semantic Computing

Tasks:

In my second research project I worked on algorithms that try to predict the political stance of short election statements. I compared both classical and deep learning techniques.

2016 - 2017: Predicting Causal Lexical Relations 

Customer: Bielefeld University, AG Semantic Computing

Tasks:

In my first research project I built a pipeline for learning causal relations (?if A then B?) in a semi-supervised way. To reach this goal I trained word embeddings on different corpora, using Gensim and NLTK.