Paul Pozorski

I study Data Science (MSc) at Warsaw University of Technology — BSc with honors. I care about making multimodal LLMs explainable and turning research into tools people can actually install and reuse.

Most of my recent work sits at the intersection of NLP, Shapley-value XAI, and practical data science. I previously built Botuj (AI chat + voice); I’m no longer actively developing it.

Roles & experience → LinkedIn

• Extending nlp-shap — Shapley-style explainability for text / multimodal models
• Preparing the ACL 2026 System Demonstration for mllm-shap
• Open to chats about XAI, reproducible AutoML, and local LLM tooling

mllm-shap: A Shapley Value Explainability Platform for Text-Audio Multimodal Large Language Models
ACL 2026 System Demonstrations — with Jakub Muszyński & Maria Ganzha

Attribution for text–audio multimodal LLMs: modality-aware masking, multi-turn conversation tracking, and phonetic audio grouping.

nlp-shap · MLLM-Shap · Zenodo · DOI · ORCID

@software{pozorski2026mllmshap,
  author    = {Pozorski, Pawe{\l} Dominik and Muszy{\'n}ski, Jakub Mi{\l}osz and Ganzha, Maria},
  title     = {mllm-shap: A Shapley Value Explainability Platform for Text-Audio Multimodal Large Language Models},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.19678283},
  url       = {https://zenodo.org/records/19678283}
}

Python PyTorch scikit-learn Transformers FastAPI SQL Snowflake Docker Streamlit