Sei sulla pagina 1di 3

Jasper Busschers

jasper.busschers@vub.be
LinkedIn : https://www.linkedin.com/in/jasper-busschers-41865b95/ (+32) 476373246

OBJECTIVE My main goal in AI is to work on projects that I believe will benefit society. I hope
to find interesting research projects that may benefit much of society on which I can
apply my skills with great passion.

EDUCATION Vrije Universiteit Brussel


Bachelor in computer science
Received in 2019

Vrije Universiteit Brussel


Master of Science in Applied Sciences and Engineering: Computer Science
profile artificial intelligence
Expected in 2020

TECHNICAL Languages : Dutch, English, basic French


SKILLS Programming Languages : Python, R, C++, C, C#, Java, Scala, Scheme, As-
sembly, Lisp, Prolog, javascript
Database : MySQL, SPARQL

Frameworks :Microsoft web stack, Azure web service , Akka play, Akka actor, Akka
streams, Pytorch, SKlearn, Tensorflow, R-shiny

Main interests : Deep learning, reinforcement learning, multi-objective optimisa-


tion, bio-informatics

EXPERIENCE Camera Assistent at VideoHouse Aug 13 - Sep 18


This job involved installing all audio and video equipment before television recordings
of both studio programs, concerts and live sport events.

contract work at Imec sept 19 - okt 19


For this job I had to go through all AI related research of Imec and create a slide deck
that summarizes these for business presentations.

PROJECTS Bachelor thesis : Computer vision based approach for analysing uids
insidetransparent glass vessels august 2018

The goal of this thesis was to create a system capable of classifying and predicting
how fluids change over time, when put in different conditions. This research was
conducted in partnership with ProcterGamble and can be fully read on the following
link (https://www.scribd.com/document/392262502/Thesis-Final).
• Technology/Tools: Python, pytorch , OpenCV

natural language processing (academic) : sentiment analysis on amazon


reviews march 2019

The first assignment in the course of NLP involved predicting the review score of
amazon reviews, based on the comment given by the user. An accuracy of 68 percent
was achieved.
• Technology/Tools: Python, Scipy, NLTK, spaCy, sklearn
• Github: https://github.com/JasperBusschers/natural-language-processing

natural language processing (academic): Semantic Frame extraction


march 2019

in the second assignment I used conditional random fields to identify verbs and the
people or objects it acts on in a text. This was trained using the colar corpus which is
a labeled frame dataset

• Technology/Tools: Python, Scipy, NLTK, spaCy, sklearn


• Github: https://github.com/JasperBusschers/natural-language-processing

natural language processing (academic): computational construction


grammar march 2019

The goal of this project was to use the fluid construction grammar that was developed
at the VUB to correctly parse meaning along with a given text. The meaning is build
up as a tree that represents the knowledge after reading.
• Technologies: LISP , fluid construction grammars
• Github: https://github.com/JasperBusschers/natural-language-processing

Advanced methods in bioinformatics (academic) : Secondary structure


prediction may 2019

The problem of secondary structure prediction is an essential problem in bioinformatics


to gain more understanding in the ways dna translate to protein structures. In this
project I preprocessed a dataset of chromosomes and tested multiple algorithms when
predicting the secondary structure.
• Technologies: sklearn
• Github: https://github.com/JasperBusschers/Secondary-structure-prediction

Multi agent learning seminar (academic) : Self imitation learning


experiments august 2019

In this project I created my own adaptation of the paper generative adversarial self
imitation learning. Here the original reinforcement learning is given another objective
to also imitate the best past experiences.
• Technologies: Pytorch, openai gym
• Github: https://github.com/JasperBusschers/Self imitation experiment

Actual trends in artificial intelligence (academic) : Hand gesture


recognition from images using modular convolutional neural networks.
august 2019

In this project I implemented 2 modular neural networks to process images of people


and classify which hand gesture they are making. The approach consisted of 2 modu-
lar networks, the first a neural network for pixel-wise segmentation of the hands. The
second network these segmentation maps then to make its prediction.

• Technologies: Tensorflow 2.0, openCV


• Github: https://github.com/JasperBusschers/gesturer ecognition

Game theory (academic) : emotional multi agent reinforcement learning


december 2018

In this group project we implemented the paper ”Emotional multiagent reinforcement


learning in spatial social dilemmas”, which proposes a simple model of emotions to
improve cooperative behaviour in multi agent environments.
• Technologies: networkx
• Github: https://github.com/JasperBusschers/emotional-multi-agent-reinforcement-
learning-on-social-dilemmas
Information visualisation (academic) : Interactive data visualisatio of air
traffic delays in US. december 2018

This course was all about creating good interactive data visualisations that allow in-
tuitive exploration of different variables. The project is still ongoing but can be shown
when requested.
• Technologies: R , R-shiny

RELEVANT • Theory of computation • Information theory • Declarative programming


COURSES • Software architectures • Open information systems
• multi-agent-learning seminar • natural language processing
• computational game theory • adaptive system seminar • advanced methods in bioin-
formatics • biotechnology • paralelism and distribution • statistical foundation of ma-
chine learning

MASTER In the year 2019-2020 I will be working on my master thesis, an idea I came up with to
THESIS use imitation learning in multi objective optimisation problems. As oppose to single
objective optimisation problems, in multi objective problems there does not exist a
unique best solution. In my thesis I will propose an architecture that is able to discover
all of these solutions, after which the end user can express which solution he prefers.

Potrebbero piacerti anche