Hi, I'm Mate Kovasznai, a 3rd year Data Science & AI student at Breda University of Applied Sciences.
My main areas are natural language processing, computer vision and data analysis. Right now I'm going deeper into the MLOps lifecycle, because I enjoy taking a model beyond the notebook: containerising it, deploying and measuring whether it actually works in practice.
Over the past few years I've built a wide range of machine learning projects. I created an end-to-end system that turns video and audio into per-sentence emotion predictions, developed a vision-guided robotics pipeline that inoculates plant root tips with a 0.5 mm error, and engineered a real-time transcription soundboard that responds to spoken words in under half a second.
I've also trained a skin-disease classifier, built a cloud pipeline that predicts NO₂ levels around Breda, and worked hands-on with the Unitree Go2-W and G1 humanoid robots using ROS2, lidar and isaac-sim.
As I mentioned right now I'm focused on production MLOps, statistics, and data storage tools and concepts, I'm looking for a data science or ML engineering internship, or a data analyst or translator role, where I can ship and measure real systems.
When I'm not in the lab or behind a screen, you'll probably find me playing chess ♟️
|
End-to-end MLOps system that turns video/audio into per-sentence emotion predictions for a TV production client.
|
Computer vision + robotics pipeline in PyBullet, reaching a 0.5 mm error in the final pipeline.
|
|
Real-time transcription soundboard: live audio → streaming ASR → regex-triggered sound effects in ~300–500 ms.
|
Tinder-style news reader with ML summaries on the back of every card. ML runs in under ~600 MB on an Oracle Always Free VM.
|
|
Behavioural cloning agent with minimap-based probability mapping of enemies. (Ongoing, in data gathering phase.)
|
Development on the Unitree Go2-W and G1 humanoid robots.
|
