Hi, I am Jan Jetze
- Data Scientist, Software Engineer &
- musician
- singer-songwriter
- looking for something new
Since I started to learn Python back in 2015, I’ve been passionate for programming. This made me graduate Cum Laude for the master Data Science in July 2018. Since then I have published a scientific paper, teached Machine Learning classes and made recommendation models for hundreds of thousands daily news consumers. And never have I ever stopped learning.
My curiousity and eagerness to learn makes me figure out new programming languages. My enthousiasm and passion for Machine Learning helps me explain its pros and cons in everyday life. My perserverance made me able to solve a Rubik’s Cube. And last but not least, my creativity helps me write down my own thoughts in a song.
In my role as a Data Scientist I have created different recommendation models. Via the Primed IO delivery-server these models serve recommendations to hundreds of thousands daily news consumers. Using tools like PySpark I had to filter user-interaction data of up to 1TB to evaluate and evolve the recommendation models over time.
At Primed IO I have learned that I can only be effective as a Data Scientist if I have Software Engineering skills too. After all, code only really matters when it is deployed into production. Due to the small team we worked in I had plenty of opportunities to develop those Engineering skills. With these skills I was able to deploy and manage all Machine Learning models on the Kubernetes cluster. Furthermore I wrote an Rest API for ab-test evaluation and middleware for Data Scientists willing to use the Primed IO service.
September 2017 - April 2019, Amsterdam
September 2018 - April 2019
In this function I mainly acted as a super TA (Teaching Assistent) which meant organizing and teaching weekly tutorials for different courses. Also I supervised groups in their half-year project.
July 2018 - September 2018
Together with the PhD student who was my daily supervisor during my Thesis project, we continued research into the topic I wrote about for my Thesis. At the end of September we submitted the resulting paper to the ICLR 2019 conference;
PIE: Pseudo-Invertible Encoder
September 2017 - April 2018
Helping and grading students with the weekly assignments they’ve got to code in C#, Python or JavaScript.
May 2013 - May 2017, Amsterdam
April 2015 - May 2017
Keep an overview of the student desk, stay in contact with supplier of bikes for the bike sale, doing direct debits for members and organising events like twice a year a bike sale with up to 400 bikes.
May 2013 - April 2015
First contact point for students in Amsterdam. Answering questions about education, keeping up-to-date the member administration and organising services like a weekly bike sale.
University of Amsterdam (UvA)Graduated July 2018, CUM LAUDE
Orthogonal matrices have already been used in Neural Networks because of their good properties such as invertibility and norm preservation. However, orthogonal matrices can be used only in mappings directly parametrized by square matrices. I present a method that combines several useful properties of orthogonal matrices and allows for dimensionality reduction at the same time. I demonstrate that this method is able to achieve similar results to comparable methods. The method I propose, however, has significantly fewer parameters and it converges to be invertible.
University of Amsterdam (UvA)Graduated January 2017
University of Amsterdam (UvA)Graduated July 2017
On November 14 2014, 5000 students gathered on the Malieveld in The Hague to protest against the Dutch Student Loan. One of the acts that day was me singing this self-written protest song.
A year after the implementation of the Student Loan, the government presented statistics showing negative results of this policy. As a means of ‘I told you so’, I recording this sequel protest song.
In the year 2014 one of my favorite musical duo’s - Acda & de Munnik - announced to go apart. I decided to send a musical open application to take over the position of Paul de Munnik. I didn’t get further than a mention by Thomas Acda though…
Submission to the ICLR 2019 Conference. We introduced a new class of Autoencoders with pseudo invertible architecture.
DetailsThis Github page shows a visualisation of the futurue of the Dutch Pension Act. This was the graduation assignment for the Minor Programming which I graduated for in January 2017.
DetailsHere you can find the thesis I wrote for the master Data Science. In this thesis I created a Neural Network layer based on orthogonal matrices. This layer was able to reduce the number of dimensions of its input while still being able to reconstruct the original input.
Details
