Neuroscience, machine learning, reinforcement learning, robotics, brain-computer interfaces

This page is a collection of my blog posts, talks, publications, ideas and projects. They touch on machine learning and computer science, neuroscience and brain-computer interfaces, reinforcement learning and robotics.

As a machine learning architect at OffWorld Inc I am leading a machine learning team to train robots for industrial applications in unstructured environments, aiming at space exploration. In my opinion, reinforcement learning has reached a point where we can attempt to deploy it on real-world robots for practical applications.

I have received my PhD from University of Tartu in 2020 for my thesis “Understanding Information Processing in Human Brain by Interpreting Machine Learning Models” that explored the intersection of Neuroscience and Artificial Intelligence. My focus is on interpreting machine learning models that were trained to decode intracerebral electrophysiological activity of human brain. For a long time humans were in the business of carefully crafting elegant models that describe their observations of reality. An accurate and insightful model reveals internal dynamics of a system and leads to knowledge. In the modern times the amounts of data to sift through became unmanageable, so the field of machine learning had emerged to look for patterns and build models automatically. However, something was lost in that transition from manual to automated modeling. And that something is the precise understanding of how exactly a model operates. I believe that interpretation of machine learned models, in particular the ones that were trained on life sciences data, is the way to get the best of the both words — an automated model-building capability that can process huge volumes of data and careful and insightful understanding of the underlying process.