2. About

Michał Chromiak, PhD

I am a computer scientist at Maria Curie-Skłodowska University (UMCS) in Lublin, Poland. My interests include machine learning, deep learning, and software engineering, with a background in integrating heterogeneous and distributed systems.

I hold a PhD in Computer Science from the Institute of Fundamental Technological Research, Polish Academy of Sciences, and master's degrees in Mathematics and Computer Science. My doctoral research focused on a unified platform for integrating heterogeneous resources.

University profile · ORCID · GitHub

Research and engineering🔗

My experience spans academic research and applied software development. Selected work includes:

  • Object databases. Contributions to the ODRA database prototype and research on the Stack-Based Approach to databases at the Polish-Japanese Academy of Information Technology during my doctoral studies.
  • Geospatial software. Development of an aerial-image georeferencing platform for the Institute of Soil Science and Plant Cultivation (IUNG).
  • Healthcare software. Design of transaction-like behaviour on top of Elasticsearch for CompuGroup Medical.
  • Computational biology. Exploratory work in 2016 on modelling CRISPR workflows using MapReduce.

I also served on the ECAI 2024 Programme Committee.

About this blog🔗

This is my space for working through machine-learning ideas and explaining the research behind them. I write about foundations as well as papers in representation learning, computer vision, language modelling, and reinforcement learning.

What draws me to the field is the way mathematics can turn observations into useful models. I am also interested in its connections with neuroscience, psychology, and philosophy. The teaching of Andrew Ng and Fei-Fei Li helped shape my interest in making machine learning more accessible.

In each article, I aim to connect the problem, the intuition, the mathematics, and the experimental evidence. I distinguish what a paper demonstrates from what its results might suggest.

These articles grow out of my own reading and learning. Corrections and thoughtful questions are welcome.

Code and learning resources🔗

I collect research implementations and learning materials in two GitHub organisations:

  • ML-Kraft: implementations and code related to machine-learning papers, including forks of third-party projects.
  • ML-Dojo: learning resources for machine-learning tools and frameworks.

For further reading, see the resources page. Familiarity with Python, linear algebra, and multivariable calculus is useful for the more technical articles; NumPy, pandas, and Matplotlib provide a practical starting point for working with data.

Get in touch🔗

For research discussions, collaboration enquiries, article suggestions, or corrections, please use the contact page.