Hey everyone, and welcome to my blog! My name is Alexey Abramov. I’m Russian, born in Moscow, and currently based in Munich, Bavaria (BY), Germany. I like building reliable and scalable software in the fields of computer vision, artificial intelligence, and robotics. I hold a PhD from Georg August University of Göttingen (2012) and diploma from National Research Nuclear University MEPhI in Moscow (2008) in computer science. I possess hands-on knowledge in modern software design and strong coding skills in Python and C++. You can find my curriculum vitae here: Resume Alexey Abramov.

In my career I have worked at several organizations on a large number of complex robotics projects ranging from self-driving cars and driver assistance systems to replicating human actions with robots. At the moment I’m working at CARIAD SE @ Volkswagen Group on behavior prediction for highly automated driving functions. Prior to that I worked on behavior prediction for self-driving cars at Argo AI (transferred to CARIAD SE in 2023). Earlier I spent 9 years at Continental AG where I was occupied with computer vision algorithms for Advanced Driver Assistance Systems (ADAS) and highly automated driving. Before joining the industry, I worked on perception for intelligent autonomous robots as part of my PhD studies in the Computational Neuroscience Group at Georg August University of Göttingen.

My first computer program was written in Visual Basic when I was 15 with the help of my dad, who was a CAD expert for circuit development with decades of hands-on experience. I have picked up the basics of structured programming with Pascal during the first semesters of my studies. In 2004 I was exposed to C++ for the very first time. I was incredibly lucky to have an amazing tutor who wrote innumerable lines of code with a chalk on a blackboard explaining every single aspect of the language. Believe it or not, in my first object detection implemented back in 2006 (rule-based only stuff) I had to write custom memory allocators for grabbed camera frames. Sliding over pixels and color channels was done with raw pointer arithmetic. Good old times! Despite all the new features introduced by the recent standards, C++ remains by a good margin one of the most difficult languages to master. Practice makes perfect, there is no way around C++ in the robotics space (at least not yet).

Around 2010 a group of talented master students in Göttingen convinced me that Python (2.5 or 2.6 that time) is the way to go when prototyping ideas in computer vision / machine learning / artificial intelligence. Today Python is my favourite programming language used almost every day. Caffe became the first deep learning framework I got my hands on in 2015 (a few years before TensorFlow and PyTorch took over the field). I contributed to one of the first demo drives using deep learning based perception in the history of Continental AG for the IAA 2017.

The technological landscape around us is evolving rapidly. There has never been such an exciting time for building intelligent machines. I’m incredibly excited for what the future holds. To keep up with the pace I spend some time in the evenings and at weekends reading technical literature, research papers, sharpening programming skills. Almost every day I’m trying to approach one LeetCode problem to not lose touch with computer science basics. From time to time I try to publish occasionally at machine learning / computer vision / robotics conferences just for fun:-)

In my spare time I do sports, read books, explore new places, attend cultural events, and, last but not least, enjoy time with my family and raise my little son:-)

Thanks for visiting my blog, I hope you will like it!