Simon Thordal

Mathematician and programmer who likes ML and complex networks

Education

2014 - 2016

M.Sc. Mathematical Modeling and Computing; Technical University of Denmark, Kgs. Lyngby, Denmark

Specializing in machine learning and complex networks I ended up writing my final thesis, Link Prediction in Complex Networks with Node Metadata, on the prediction of citations made by the verdicts of the European Court of Justice.

2009 - 2012

B.Sc. Mechanical Engineering; Technical University of Denmark, Kgs. Lyngby, Denmark

My first meeting with engineering where I spent more time in Matlab than I did designing mechanisms which led me to mathematics and my bachelors thesis on digitally reconstructing 3D-printed items to make it possible to measure internal dimensions.

Work experience

2013-2015

Junior Developer; 3Shape, Copenhagen, Denmark

Part time job next to my studies working on developing interfaces between our CAD software and 3rd party milling machine software as well as verification of milling machines. Mainly worked in C# and Delphi and got experience working in a small scrum team in an enterprise with 300 developers and millions of lines of legacy code.

2012-2013

Web Developer; Lagoa (Autodesk), Montreal, Canada

Internship turned full-time job where I worked in web development for a start-up performing browser based 3D rendering. Mainly worked with JS/RoR and Redis and I got a lot of experience with the fast turn-arounds in the start-up world as well as a baptism by fire in web development.

2009-2012

Translator; Self-employed

Worked as a translator from English, Swedish and Norwegian to Danish.

2007-2009

Partner manager / Customer agent; Scalepoint, Copenhagen, Denmark

Worked first as a customer service agent and afterwards negotiating supplier deals in a large network of Norwegian suppliers of all things from cell-phones to traditional folk dresses. High stress environment with some interesting challenges.

Projects

• Link prediction in complex networks My master thesis focuses on using machine learning tools to predict future links in a network of European case law verdicts, useful both as a recommender engine and a way to infer network structure.
• Topic clustering in EU plenary debates I used dimensionality reduction (ICA) and unsupervised learning to mine topics in EU plenary debates to try to cluster parties together based on their favoured causes as well as what their stances were.
• From scratch neural networks implementation To get an in-depth understanding of how neural networks work I created one from Python built-in functions. It is slow, but as an educational tool it works well.

Technologies

Coding techniques and tools:

• Python and some of its machine learning eco-system (sklearn, pandas, networkx etc)
• Matlab
• C# in an enterprise setting
• Neo4j
• Used to writing production level, OOP code
• Familiar with the general web development architecture (REST, MVC, SQL / NoSQL databases)