Session Outline

We would like to share our experiences, thoughts, and results on partial automation of car damage reports. The presentation will describe our journey in the past, present and the future regarding business idea, technical aspects, challenges, and solutions.

Key Takeaways

  • Go slow to go fast when carrying out a large project.
  • Break problem in less complicated pieces rather than trying to solve everything at once.
  • Data heterogeneity is a nightmare.
  • AI has hard time learning certain things, such as human experience and hunch.


Speaker Bio

Vladimir Tripkovic – Data Scientist | Topdanmark

I am educated as a chemical engineer from the University of Belgrade, Serbia. I have though lived the last 15 years in Denmark, where I first did my PhD in Physics at the Technical University of Denmark and then spent several years in research in various academic groups. My worked was centered around computational catalysis with almost no connection to machine/deep learning. However, with time I grew tired of academia and wanted to make a carrier change into the private sector. I saw machine/deep learning as an obvious place to start, decided to invest in it and have not regretted it ever since. After spending ca. 9 months on online courses and books I got the opportunity to work for Topdanmark. There, I have spent the last three years working on computer vision problems involving cars and buildings as well as on improving risk assessment parameters.

November 9 @ 13:50
13:50 — 14:20 (30′)

Day 2 | 9 Nov 2022 | MACHINE LEARNING + MLOPS

Vladimir Tripkovic – Data Scientist | Topdanmark