Jan von Bassewitz
vbjan[at]ethz[dot]ch
Data Science MSc, Mechanical Eng BSc
About Me
Hi I’m Jan. I believe in the beauty 🌟 and power 💪 of mathematics. By teaming up with like-minded people 👥, I love to translate math into code 💻 to find intelligent solutions for real world problems.
This mission led me to study Data Science for my Master’s at ETH and Harvard, which I finished at the end of 2023.
During two research projects at these institutions, I worked on problems in computer vision and scientific computing:
-
🌊 Throughout my 9-month full-time research stay with Petros Koumoutsakos at Harvard, I tackled the problem of data-driven turbulence models in computational fluid dynamics. To that end, I developed a new method to learn such a turbulence model using reinforcement learning. The resulting paper is currently under review. You can find the preprint here [Link].
-
🌳 During my 3-month research project at Stefano Mintchev’s lab at ETH, I worked on depth estimation solutions to infer occluded tree branches from RGB-D imagery. The resulting paper is accepted in IEEE RA-L [Link].
Currently, I am actively seeking a PhD position and am particularly excited about representation learning and scientific ML.
If this text resonates with you, feel free to reach out. I love meeting new people. You can reach me at vbjan[at]ethz[dot]ch.
Publications
You can find an up-to-date collection of my publications on Google Scholar.
Projects
Autograd: [Link]: Automatic differentiation and gradient descent from scratch for arbitrary optimization problems.
Tree Branch Detection [Link]: Depth estimation of leafless branch structure given RGB-D image of tree using CNN.
Bachelor Thesis on Neural ODEs [Link]: Depth estimation of leafless branch structure given an RGB-D image of a tree using CNN.
Gaussian Processes for Control [Link]: Studying Gaussian Processes and writing a comprehensive report on methods of propagating uncertainty for control purposes.
Webscraping Bot [Link]: Automation of gym reservations.
2D-FEM on L-shape Domain [Link]: Numerical approximation of the stationary reaction diffusion equation on L-shape domain.