Hi, I'm Witek! ChatGPT said that I am a mix of coder, storyteller and maths enthusiast constantly refining both my technical skills and my worldview.
Well, who am I to argue with AI
But in all honesty, I do enjoy learning new things in the worlds of math and computer science. I love writing code, and I’m a little afraid I might be slightly addicted to JetBrains IDEs. I enjoy solving problems — whether it’s through writing efficient algorithms, optimizing systems, or just making something work better than before. When I’m not coding, I’m probably reading about AI, working on a writing project, or hanging out with friends… so I don’t end up talking to my computer screen. 😅
Always open to interesting discussions, collaborations, and learning from others!
Currently working at Capgemini:
Roles:
- Data Scientist
- Software Developer
Commercial experience in:
Gates Open Research (link)
Role: Data Scientist
Description: Automation of histological analysis of onchocerciasis nodules with artificial intelligence.
Global Data Science Challenge (GSDC) 2024
Role: Lead Data Scientist
Description: I was one of the organizers of GDSC. It is an annual, purpose-driven hackathon that brings together thousands of participants globally to tackle real-world challenges using AI.
2024 edition focused on agentic AI systems, and in cooperation with UNESCO, tackled problems in worldwide education and reading skills.
Role: Data Scientist, Researcher
Description: I worked in a small team on testing how different methods of text embedding influence the quality of retrieval augmented generation (RAG). Publication in review.
Artificial Inteligence major at Wrocław University of Science and Technology (PWR) (2023-2024).
Master's Thesis: Grammatical embedding of texts and its applications in clusterization, visualization and classification of texts in Polish
Abstract
The work investigates the effectiveness of grammatical embeddings of texts (grembeddings) in the tasks of clustering, classification, and visualization of texts in Polish. Grembeddings, unlike semantic vectorization, represent texts based on their grammatical structure, capturing syntactic and morphological features. The study explores the application of grembeddings in combination with various traditional and deep learning-based vectorization techniques, including Bag-of-Words, TF-IDF, and variants of BERT. The research utilizes seven diverse Polish datasets, encompassing literary texts, newspaper articles, social media posts, and fanfiction stories, with varying levels of semantic similarity and authorial styles. The experimental results demonstrate the effectiveness of grembeddings in author identification tasks, especially for texts with similar semantics but different writing styles. In tasks requiring thematic analysis, semantic embeddings outperformed grembeddings, highlighting the importance of semantic information in clustering and classification. The results suggest that grembeddings are a valuable complement to traditional text representation methods, particularly in capturing stylistic nuances and enhancing author attribution.Applied Computer Science at Wrocław University of Science and Technology (PWR) (2019-2023).
Bachelor's project: A multilevel, multiplayer educational game, focused on explaining the genetic algorithm and its implementation in Python (link).
Additional skills learnt during personal projects:
Qwery List is a small library introducing a new way to use higher order functions with lists, with lazy evaluation. It allows to chain well known python functions such as map, filter and reduce. For further details visit the PyPi page or the official documentation.
Simple commandline application for converting mp4 movies to videos of frames made out of ascii characters (link)
Implementation of minmax algorithm to play classic checkers. Code written purely in Rust with multithreading optimisations and alpha-beta pruning optimization.