Created during a so called "experimental week", Toptracer Access was a product to accompany the Tournament Range suite that Toptracer offered. Its goal was to allow anyone anywhere to see professional players' shots at tournament range events (compared to it only being possible on LED screens at the event previously).

The idea for this product was conceived by me, and it was designed and developed by me and two awesome colleagues. During a second experimental week, it was also deployed to access.toptracer.com, but this doesn't seem to be the case any longer.

When I bought my car, I thought it might be useful for friends or family to have a way of calculating the costs for borrowing it. Therefore, I built a website for doing exactly that!

You can find the webpage at brum.omstedt.com and find the source code here.

During the spring of 2020, I wrote my Master's Thesis at Topgolf on the topic of deep reinforcement learning together with golf. I investigated whether a deep reinforcement learning agent could utilize shot data from a real golfer in order to learn how to play golf. This was done to figure out if a virtual entity could learn more about the shot data than the golfer already knew, i.e. learn how to play golf better than the golfer. If so, the agent could then act as a virtual caddie telling the golfer how to best play golf given their shot characteristics.

My research concluded that it is possible to utilize deep reinforcement learning with golf. However, it was inconclusive in regards to whether an agent could play golf better than a golfer, since the agents I trained only managed to play on par (pun intended) with the golfer. There were several possible reasons behind this, so more research should be conducted to determine the usefulness of reinforcement learning in golf.

If you are interested in learning more, you can read the thesis here.

For my ex-girlfriend's birthday, I decided to create a game in which Hanna has to save Fredrik (me) from the evil Dr. Demon and his minions. The game was written using GameMaker Studio 2, and everything from graphics to code was created by me.

The source code for Hanna is available here.

In the spring of 2019, I joined the band Paper Plane Pilots to play the guitar for a gig hosted by Spotify at their office in Stockholm. It was a lot of fun and took many weeks of practice. The entire show is available in this video.

As part of a course on AI at KTH, I was tasked with making a car traverse maze structures as fast as possible from a start position to a goal position. The solution utilized a Rapidly-exploring Random Tree (RRT) with a kinematic motion model and a controller that handled movement. Out of the groups in the course, my group's solution performed very well, finishing in the top 3 in regards to all terrains tested. More info can be found here.

During high school, a couple of friends and I started our own video game company called Valkyr Games. Although a fun experience, it was never very successful. A few years ago, we decided to call it quits, having learned a lot about what it takes to run a company.

During Valkyr Games' lifetime, we published two apps on the App Store and Google Play. These can still be found on Google Play.

Aside from the projects mentioned above, I also have a GitHub containing some of my work (including this website). It can be found here.