# Maj Korent

> Software Engineer & Computer Vision Researcher - I build systems that have to survive contact with real hardware, real deadlines, and real production traffic.

I design and ship software that has to work outside a demo: production platforms at WMinus, tiny-footprint vision models running on competition robots, and research code that gets checked against real datasets, not slideware. Currently finishing a Computer Science degree at the University of Maribor while working across full-stack engineering, embedded systems, and applied ML.

## About

I'm a software engineer and computer-vision researcher based in Maribor, Slovenia, currently completing a Bachelor's in Computer Science at the University of Maribor. My work sits at the intersection of practical engineering and research rigor - I care whether a model still works when it's running on a microcontroller with less than 1 MB of memory, not just whether it works in a notebook.

Since September 2022 I've been building production systems at WMinus, a software studio, where I've worked across full-stack web development, computer vision, and internal tooling - including an ERP platform with tax-certification and localization for Slovenia and Austria, integrations with industrial robots and CNC equipment, and the CI/CD and architecture that keeps other engineers' deployments boring.

On the research and robotics side, I'm a long-standing member of Slovenia's national robotics team (SERS Team), where I work on the computer-vision and embedded systems that decide whether a competition robot actually performs. That includes a custom character-detection model that runs at roughly 120 FPS with about 90% accuracy in under 1 MB of memory, and hands-on STM32 firmware work: custom drivers for IR sensors, color sensors and motors, plus contributions to the team's PCB design. In parallel, as a Young Researcher at the Faculty of Electrical Engineering and Computer Science Maribor, I've optimized image-recognition models (about 5% fewer false detections in real-world testing), processed and labeled 9+ TB of image data, and published research on machine-learning techniques.

I'm comfortable across the stack this implies: C++ and Python for embedded and ML work, TypeScript/JavaScript for web platforms, ROS/ROS2 for robotics integration, and the cloud and CI/CD tooling (Docker, Cloudflare, AWS/GCP/Azure) that ships all of it. I also build my own developer tooling - agent-session dashboards, trace/memory systems, and automation - because if a workflow is annoying twice, it's worth automating.

## Experience

### Lead Software Engineer - WMinus (Oct 2025-Dec 2025)

Led software engineering work spanning SaaS development and delivery. Owned technical strategy and architecture decisions across client and internal projects.
- Architected and implemented scalable solutions across client engagements
- Coordinated technical strategy and team workflow

### Computer Vision Engineer - WMinus (Jan 2025-Oct 2025)

Developed and deployed computer-vision systems for client and internal use, from model training through production integration.
- Implemented deep-learning models for image processing, object detection and analysis
- Built and maintained CV pipelines with PyTorch and OpenCV

### Web Developer - WMinus (Sep 2022-present)

Ongoing production engineering at a software studio: client web platforms, internal tooling, and a multi-tenant ERP system with tax certification and SI/AT localization, including industrial robot and CNC integrations.
- Built roughly 50 websites and applications for clients
- Built internal tools that cut client project costs by about 10% on average and shortened delivery time by roughly a third
- Contributed to an ERP system with tax certification and Slovenia/Austria localization, including industrial-robot and CNC integrations
- Maintain architecture and CI/CD pipelines across projects

### Software Engineer (Internship) - SSRD d.o.o. (Nov 2025-Jan 2026)

DevOps and embedded-software work on a secure UAS (unmanned aerial systems) project. Details are limited by an NDA on the underlying project.
- DevOps for secure embedded software in UAS
- Built a camera pipeline for calibration and diagnostics
- Worked with ROS2 Humble

### Computer Vision Researcher (Young Researcher) - Faculty of Electrical Engineering and Computer Science, Maribor & SERŠ (Sep 2022-present)

Applied computer-vision research: CNN architectures for image classification, large-scale dataset processing, and published research.
- Optimized large-scale image-recognition models, cutting false detections by about 5% in real-world testing
- Processed and labeled 9+ TB of image data with consistent validation systems
- Built a supervised pipeline for neural-network training data, reaching over 80% accuracy on a 100GB dataset
- Trained CNNs based on VGG-style architectures (Conv2D + MaxPool, dense head, dropout) for binary image classification
- Published research on machine-learning techniques and presents regularly at young-researcher meetings in Slovenia and internationally

### Computer Vision / Machine Vision Engineer - SERS Team Robotics - Slovenian National Robotics Team (Sep 2023-Nov 2025)

Computer-vision and embedded-systems engineering for Slovenia's national robotics team, competing at RoboCup and related international events.
- Built a custom character-detection model running under 1 MB of memory at a stable ~120 FPS and ~90% accuracy
- Developed STM32 firmware for the robot platform: custom drivers for IR sensors, color sensors, motors and internal communication
- Contributed to custom PCB design for the robotics platform
- World finalist, RoboCup 2025 - 3rd place in the technical challenge

### .NET Developer (Internship) - ZENN - Web Solutions (Oct 2023-Nov 2023)

Web application development using the .NET framework and C#, in an agile team environment.
- Developed web applications using .NET and C#
- Worked within agile development practices

## Projects

### RoboCup Rescue Vision Model (2025)

Lightweight victim-recognition model for RoboCup Rescue Maze, running real-time on a microcontroller-class device.
- Tech: OpenMV, FOMO, Computer Vision, STM32, Embedded ML
- Repo: https://github.com/korentmaj/RoboCup-Rescue-Model-2025

### Rescue Maze Simulation (Erebus, 2024 season) (2024)

Full simulation environment for the 2024 RoboCup Junior Rescue season, used to test navigation and detection strategy before hardware deployment.
- Tech: Python, Webots, Simulation, Robotics
- Repo: https://github.com/korentmaj/RcupSimulation2024SloveniaNational

### Galaxy Classification Study (2025)

Research pipeline for classifying galaxies with modern machine-learning techniques.
- Tech: Machine Learning, Python, Astronomy
- Repo: https://github.com/korentmaj/Classification-of-Galaxies-Using-Machine-Learning

### Bipropagation Study (2026)

Independent reproduction and component-level breakdown of Bojan Ploj's bipropagation training method, benchmarked against modern backprop.
- Tech: Python, Deep Learning, Research
- Repo: https://github.com/korentmaj/bipropagation-study

### Freeflow (2026)

Press a hotkey, dictate, get clean text back. Works in any app.
- Tech: Swift, macOS
- Repo: https://github.com/korentmaj/freeflow

### mnemosyne (2026)

Exports and anonymizes coding-assistant conversation histories into a unified dataset.
- Tech: Go
- Repo: https://github.com/korentmaj/mnemosyne

### Agent Session Dashboard (2026)

A dependency-free local dashboard and TUI for monitoring AI coding-agent sessions.
- Tech: Python, TUI
- Repo: https://github.com/korentmaj/simple-agent-session-manager

### XRPL Bridge Monitor (2025)

Wallet-monitoring tool for the XRP Ledger with transaction tracking and watchlists.
- Tech: JavaScript, XRPL
- Repo: https://github.com/korentmaj/XRPL_Bridge_Monitor

### Industrial ERP & POS Platforms (2026)

Production ERP and point-of-sale platforms for Slovenian manufacturing and retail clients, with tax certification and multi-country localization.
- Tech: TypeScript, Python, Odoo, CI/CD

## Skills

- **languages:** C++, Python, TypeScript/JavaScript, C#, SQL
- **frontend:** HTML/CSS, Vue
- **backend:** Node.js, .NET, Odoo/Python
- **ai:** TensorFlow, Keras, OpenCV, CNN architectures, TinyML / edge inference (FOMO)
- **infra_cloud:** Docker, CI/CD, Cloudflare, AWS, Google Cloud, Azure, Ubuntu, Apache
- **embedded_robotics:** STM32, ROS/ROS2, I2C/SPI/UART, PCB design, OpenMV, Webots
- **tools:** Git, GitHub, Adobe Photoshop/Illustrator/Premiere, AI coding-agent workflows (Claude Code)

## Education

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

### Secondary Electrical/Computer Education - School of Electrical Engineering and Computer Science Maribor (SERŠ) (2021 - 2025)
- President of the school student council
- Institute council member; Student Organization of Slovenia (DOS) activities
- Collaborated with Slovenia's Ministry of Digital Transformation on digital initiatives
- Top ranks in multiple national science competitions

## Awards

- RoboCup 2025 World Championship - 3rd place, technical challenge - RoboCup (2025)
- Student of the Year distinction - Elektro Maribor (2025)
- Gold award, Young Researchers - Young Researchers Slovenia (2024)
- CS50: Introduction to Computer Science - Harvard University (edX) (2020)
- Data Science: Machine Learning - Harvard University (edX) (2020)
- Entrepreneurial Skills Pass - Junior Achievement (2023)

## Contact

Email: maj.korent@sers.si
Open to discussing collaboration, robotics/CV work, and select engineering projects.

CV downloads: [en](Images/doc/MajKorentCV_en-US.pdf), [sl](Images/doc/CV_SLO.pdf), [de](Images/doc/MajKorentCV_de.pdf), [lb](Images/doc/MajKorentCV_lb.pdf)