# About - Maj Korent

An engineer who builds robots, the vision that runs on them, and the agent infrastructure that ships everything else.

## About

I started with machine-learning research in 2020, because a robot needed to see something and nothing off the shelf did it. That turned into Slovenia's national RoboCupJunior team, a 3rd place at the 2025 world championship, and the SuperTeam title in Korea in 2026.

Day job: computer vision at WMinus since January 2025, robotics engineering there before that. On the side I consult for companies I hold a piece of, I have built drone software, ERP systems with tax certification and industrial robot integrations, and roughly fifty websites and apps that are somebody else's problem now.

I do too many different things. I know. The upside is that the hardware, the model and the deployment are usually the same person's fault, so they get fixed the same afternoon. Whatever turns out to be useful to anyone else goes on GitHub for free.

## Away from the keyboard

- **Wakeboarding:** Behind the boat all summer. Falling is most of the sport.
- **Sport driving:** Apexes, late braking, and one more lap than sensible.
- **Hardware:** Iron on, board out, smoke occasionally. The best kind of debugging.
- **Sailing:** Wind, a sail, and nothing that needs a deploy.
- **Painting:** Slow, analogue, and nobody reviews the diff.
- **Photography:** Mostly light and water, occasionally a robot.

## What I work in

- **Languages & Web:** Day to day: TypeScript and JavaScript for the web and mobile work, Python and C++ for anything closer to hardware or a model, C# and SQL from jobs that needed them, Vue and plain HTML/CSS on the frontend, and Node.js, .NET or Odoo/Python on the backend depending on whose stack I inherited.
- **Computer Vision & Machine Learning:** For computer vision and ML: TensorFlow, Keras and PyTorch for training, OpenCV for anything that touches a live camera feed, CNN architectures as the default starting point, and TinyML techniques like FOMO for squeezing a model onto a microcontroller instead of a GPU.
- **Embedded & Robotics:** On the robotics and embedded side: STM32 firmware, ROS and ROS2, I2C/SPI/UART drivers, OpenMV for on-device vision, Webots for simulation before touching hardware, and enough PCB design to be dangerous.
- **Infrastructure & Agent Tooling:** For infrastructure and everything that keeps it running: Docker, CI/CD and Cloudflare in production, AWS, Google Cloud and Azure depending on the client, Git and GitHub for everything, and Claude Code and the Claude Agent SDK, plus custom MCP tooling, running the scheduled and autonomous agents that keep several codebases moving without me watching all of them at once.

## Education

### Diploma, Artificial Intelligence - Technical University of Munich (2025)
- Diploma of AI from Limits of Artificial Intelligence, in partnership with Alma Mater Europaea

### Bachelor's Degree, Computer Science - University of Maribor (In progress)
- Software development, Java and applied computer science

## Awards

- RoboCupJunior World Championship 2026 (Korea): 1st place, SuperTeam Challenge - RoboCupJunior (2026)
- RoboCup 2025 World Championship: 3rd place, technical challenge - RoboCup (2025)
- Gold award, Young Researchers - Young Researchers Slovenia (2024)

---

Every page: [Free resources](https://korent.dev/free/) | [Work](https://korent.dev/work/) | [Contributions](https://korent.dev/contributions/) | [About](https://korent.dev/about/) | [Frequently asked](https://korent.dev/faq/) | [Contact](https://korent.dev/contact/)