$$ \arg\min_{x,y,z}$$ Jonas Köhler

about | publications | contact



I am a machine learning researcher, engineer and consultant with experiences in Bayesian deep learning, information theory, equivariant group convolutions and natural language processing.

Currently, I am doing a research internship on deep learning under privacy constraints under supervision of Mijung Park at the Max-Planck-Institute for Intelligent Systems in Tübingen, Germany.

In parallel, I am finishing my MSc in Artificial Intelligence at the University of Amsterdam during which I did research on \( SO(3) \)-equivariant convolutions for spherical signals together with Taco Cohen in the group of Max Welling.

Before I graduated with a BSc in Applied Computer Science at the Bauhaus-University Weimar, where I worked as a research assistant on natural language processing, argumentation mining and information retrieval in the Webis group.

Prior to that I spent some time to study mathematics + physics (3y), economics + political science (1y) and new media art (sound/interfaces/visualization, 2y).

Professional Experience:

I co-founded the machine learning research engineering company Amsterdam Machine Intelligence (dissolved) during my master studies, which designed and engineered algorithms and models for startups and local municipalities.

Prior to that I worked as a software engineer intern for Betterspace in Kassel on machine learning algorithms for energy peak load prediction and a reactive microservice system for a distributed network of sensors and actors in the smart building context.

During my undergraduate studies I worked as a part-time software engineer for kuzo-media in Weimar on e-commerce systems for small to large sized companies in Thuringia.

Furthermore, I worked on many freelance and small scale projects, ranging from algorithmic market modeling for short-term energy positions traded at the EEX, over software engineering for media technology companies, to creative technology projects in collaboration with artists and designers.