Welcome!
My name is Tom Janssen-Groesbeek. In 2020 I finished my MSc in Data Science at the Radboud University in Nijmegen after which I started my first job as a Data Science Software Engineer at Itec. On this website you will find more information about my education, previous work experiences and my programming projects. Information found on this page is also summarized in my resume, which can be downloaded by clicking on the 'resume' button below.
Education
- 2016 - 2020 Master Degree Programme in Data Science, Radboud University, Nijmegen
- 2013 - 2016 Bacherlor Degree Programme in Artificial Intelligence, Radboud University, Nijmegen
- 2012 - 2013 Propaedeutic Degree in Artificial Intelligence, Radboud University, Nijmegen
- 2006 - 2012 International Baccalaureate (IB) English Programme at Kandinsky College, Nijmegen
- 2006 - 2012 Pre-University Secundary Education at Kandinsky College, Nijmegen
GPA: 7.5
GPA: 7
GPA: 7.5 (Bene Meritum)
GPA: 5/7
GPA: 7.4
Work Experience
- January 2023 - Now Scientific Programmer at iCIS, Radboud University, Nijmegen
- June 2022 - Now Tutor at Studentvoorstudent, Nijmegen
- January 2021 - Augustus 2022 Data Science Software Engineer at Itec, Nijmegen
- September 2019 - September 2020 Student-Assistent at the Donders Institute for Brain, Cognition and Behavior
- March 2019 - June 2019 Intern Data Science at the Department of Medical Microbiology at the RadboudUMC
- December 2018 - February 2019 Student-Assistent at Radboud University
- February 2016 - September 2017 Student-Assistent at Max Planck Institute for Psycholinguistics
- March 2013 - Augustus 2015 Tutor at Bijlesnetwerk.nl
Supporting researchers in cybersecurity and AI by designing, constructing and rolling out experimental prototypes. Moreover, I help prepare tutorials for students and organise lab materials.
Helping students with programming in Python and Data Science related projects.
Contribute to the software architecture and functionalities of factory automation tools and participate in the smart manufacturing group to increase production output.
Assisting with the development of the Vowel Space Travel iPad app.
Utilizing electronic patient data in order to develop a model to predict if new patients in the ER carry antibiotic resistant bacteria.
Assisting high school students with their work for the NLT course dedicated to robotics. NLT is a cource available at several high schools in Nijmegen. During the robotics module, the students learned to program a NXT lego robot to perform several tasks.
Setting up EEG/MRI experiments, guiding the subjects, analyzing speech data, and building Automatic Speech Recognition models with HKT and Python.
Tutoring high school students in math.
Projects
- Master Thesis Topic: Re-Ranking BERT; Revisiting Passage Re-Ranking with BERT on MS MARCO
- Artificial Sommelier For this project I worked together with Max Moons and we together trained a Support Vector Machine in order to classify to which score group a wine belonged based on the corresponding wine description written by a wine expert. We used several features like Bag of Words feature vectors, Latent Dirichlet Allocation topic distributions and also custom Word2Vec word embeddings.
- Weather Chatbot For this project I worked with Max Moons and we implemented a Telegram based chatbot. The chatbot could answer questions related to the weather. Moreover, the mood of the chatbot was dependend on the weather in Nijmegen. The chatbot was written in Python.
- Crawling Reddit For this project I helped a friend out with his master thesis on the relationship between social media sentiment and the fluctuations of three cryptocurrencies. I performed a web crawl where I crawled user comments on different subreddits dedicated to cryptocurrency fluctuations and speculations. Then I performed several Natural Language Processing methods to preprocesses the data and finally I performed a sentiment analysis.
- Left Lung Fissure Detection In a team with 5 fellow students we were tasked to train a neural network to detect the major fissure located in the left lung. The data consisted out of 3D lung scan images from the Radboud UMC. Our final solution was to train a 3D U-net.
- Latent Dirichlet Allocation used on Reddit Comments As final project for the course "Text Mining", I studied the correlation between online user comments on reddit and Bitcoin fluctuations using Latent Dirichlet Allocation algorithm. This algorithm is used to model the topic distribution of the collected user comments. The topic counts (mentions) and bitcoin price fluctuations were analysed for any correlation.
- Natural Computing Genetic Algorithms For this project I worked together with Tanja Crijns and Joris van Vugt to use genetic algorithms to reconstruct an image from another image. Parts of different sizes are taken from an input image to reconstruct as best as possible the target image.
- NOAA Fisheries Stellar Sea Lion Population Count Together in a team with 5 fellow students, I participated in a Kaggle challenge were the task was to write software that would automatically count the different sea lions on aerial images. We applied deep learning techniques in combination with clustering methods to the problem and ended up 26th out of 385 submissions (silver medal).
- The Nature Conservancy Fisheries Monitoring As part of a team consisting out of 5 fellow students, I participated in the Kaggle challenge where we had to write algorithms that would automatically detect and classify species of tunas, sharks and more that fishing boats catch. We applied deep learning methods to classify the fish and ended up 160 out of 2293 (bronse medal).
- Bachelor Thesis Title: Extended Lock-in Feedback Applicable on Higher Dimensional Function Maximization
Supervisor: prof. dr. ir. A.P. de Vries (Arjen)
My thesis focuses on re-evaluating BERT on its performance on passage re-ranking using the MS MARCO datset. Currently, the MS MARCO dataset contains only a few relevant passages per query, mostly only one. My hypothesis is that the dataset as is, is inaccurate and that in fact the dataset contains more relevant passages per query.
Thesis: Link
Supervisor: prof. dr. M.C. Kaptein
In this thesis I adapt the existing Lock-in Feedback algorithm. This algorithm is a means of performing stochastic optimization and the alterations should make it applicable on higher dimensional function maximization problems.