Hi, I'm Stefan
Get in touch via mail: stefan.krsteski [at] gmail [dot] com
I am a machine learning researcher and engineer from North Macedonia, working on language models, reinforcement learning, and agent evaluation. I'm naturally curious and always looking to better understand how the world works.
Currently working on RL environments for post-training and evaluation. Previously I was a research assistant at the Data Science & AI Lab, EPFL. Some things I've been working on recently:
- Long-horizon reasoning, web, and tool-use reinforcement learning environments
- AI safety specifically at the pretraining stage (aka model raising)
- Behavior simulation with LLMs and statistical inference
- Multimodal document parsing and retrieval-augmented generation
- Deploying language models for governmental and civic applications