Wrote a framework to differentiate through physics, showing that this makes training deep learned controllers for robotics remarkably fast and straightforward.
Awarded a best poster award at the Belgian-Dutch Machine Learning Conference (Benelearn).
Won $50,000 for a second place in Kaggle's Data Science Bowl with a deep neural network for diagnosing heart diseases.
Won $100,000 in a Kaggle competition with the lab team using deep learning to classify plankton.
LSD-net featured in New Scientist, Guardian, Der Spiegel, and Atlas Obscura.
Developed an interactive system for visualizing high-level features of a large-scale deep neural net.
Research on transfer learning for legged robots was published.
Successfully defended his research on the Oncilla robot to the European Commission.
The Oncilla robot achieved autonomous walking.
Research Experience
Advanced to the second round of the Facebook Hacker Cup and received a T-shirt.
Optimized a computer vision neural network by backpropagating through a camera.
Presented a poster on differentiable physics engine at the Deep Reinforcement Learning Workshop at NIPS.
Invited as a technology expert to speak with designers and journalists at the Design Museum in Helsinki, Finland.
Participated in multiple research projects such as a real-time note onset prediction system (patent pending), quadrupedal robot gait control, etc.
Successfully presented Oncilla robot research to the European Commission in Bielefeld.
Published a paper on trotting and turning techniques at Robio conference.
Background
Research interests include machine learning, deep reinforcement learning, computer vision, etc.; has extensive research experience in the field of robotics.
Miscellany
Active in Pirate Party Belgium.
Camped on the roof of an arms factory protesting against weapons export.
Spoke at the 3rd Data Science Ghent Meetup on how to use deep learning to win Kaggle competitions.
Gave a talk on the relationship between AI and Art at iMal in Brussels.