Recent Updates (see News)
About
I am a machine learning researcher. My research focuses on Continual Learning (also known as Continuous or Lifelong Learning). I am especially curious about advanced and unconventional deep learning models and derivative-free optimizers. In particular, I extended Hinton's Forward-Forward algorithm to large CNNs and am currently exploring how to adapt it to transformers. I am fascinated by how autonomous agents make decisions in the face of new or imperfect information. I also enjoy encountering intelligent artificial friends and foes in video games. I collaborated with ESA, LSA and CERN.
I recently completed my PhD in machine learning at the Polish Academy of Sciences, supervised by Prof. Paweł Morawiecki, and I am currently seeking postdoctoral and research positions.
The longer version — my research path and collaborations (click to expand)
During the summer of 2025, I was a researcher at the University of California, Santa Cruz, as part of the Summer of Reproducibility program, where I explored Kolmogorov–Arnold Network layers as structured replacements for feedforward blocks in small language models. Also in 2025, I joined a collaboration applying continual learning to data-quality monitoring for the CMS experiment at CERN. In 2024, I worked as a researcher at FDL Europe, where I adapted large-scale foundation models to work with SAR data in collaboration with the Luxembourg Space Agency and the European Space Agency. I was honored to receive two EU Horizon grants, TAILOR and ENFIELD, which allowed me to lead independent research projects on experience replay and parameter-efficient continual learning. The TAILOR grant brought me to the University of Pisa's Pervasive AI Lab, where I worked with Prof. Vincenzo Lomonaco.
I hold an MSc degree from Jagiellonian University, where I was co-supervised by Prof. Jacek Tabor and Dr. Stanisław Jastrzębski.
If you have any questions or are interested in collaborating, feel free to email me!
When it comes to coding, I prefer Python for its simplicity and the beauty of Zen. I strive for referential transparency and adhere to the single-responsibility principle with minimal side-effects, but only when it makes coding and reusing the code for other projects faster and cleaner. The spiritual side of my work is based on the Unix philosophy.
The Distilled Manifesto: Academic freedom comes first.
I support the Slow Science Movement.
Visualizations
Optimization methods:
- Original Hinton's Forward-Forward Algorithm
- My Scalable Forward-Forward Algorithm
Research interests
I have published research on the following topics:
- Continual learning: the topic of my doctoral dissertation
- Transfer, few- and zero-shot learning
- Representation learning
- Neural network optimization algorithms (Hebbian, Forward-Forward, SFF)
Work in progress:
- Reinforcement learning
- Meta-Learning
- Anomaly detection for continual learning
- AI in video games, especially agents that interact with the player (NPCs and companions). I am also interested in designing "wild" wildlife for games like No Man's Sky.
I would be excited to collaborate on projects involving:
- ML for magic tricks 🪄🎩
- Development of video games and simple text-based games (I like this PhD Simulator)
- Virtual (research) assistants, artificial scientists
- Space exploration
Of course, I have deep faith in the supreme power of evolutionary algorithms. Just check out this naturally occurring bacterial flagellar motor that spins up to 30,000 rpm with 80-100% efficiency and can reverse its rotational direction in as short as 2–5 milliseconds!
Credits: Singh, Prashant K et al. “CryoEM structures reveal how the bacterial flagellum rotates and switches direction.”
Різкими зламами, лихими закрутами
Продирається доля крізь хащі.
Хай груди заквітнуть червоними рутами –
Не втратити честі своєї нізащо!
Там, де з рідними-милими,
Там, де зі своїми хорошими,
Власним розумом, власними силами
Щастя своє збудувати зможемо!
Credits: @neivanmade, PhD