Skip to content
View patrykkozuch's full-sized avatar
8 followers · 2 following

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Shark

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Shark

Highlights

  • Pro

Block or report patrykkozuch

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
patrykkozuch/README.md

Patryk Kożuch

AI/ML Engineer | M.Sc. Student @ AGH Kraków

AI/ML Engineer and Master's student specializing in Artificial Intelligence and Machine Learning at AGH Kraków. Highly interested in new technologies, especially Large Language Models and MCP.

Key Projects

I enjoy taking on complex challenges, particularly those involving production-level deployment of novel AI techniques.

  • Qwen Recipe Generator: Fine-tunng GRPO Qwen3-0.6B model (using TRL) focused on structured output (XML) to generate culinary recipes.
  • ChatterCast: Chatterbox-Turbo based podcast generation
  • Ribbon Excercise Detection: Detecting excercises with ribbons using binary mask through 3D convolutions
  • Tokenization benchmark: Benchmark showcasing how tuning the tokenizer to the dataset used affects the training.
  • Evaluating Frame Stacking Impact in RL: Diploma project for Bachelors degree - evaluates both Frame Stacking and Recurrent Neural Networks in Reinfrocement Learning training
  • Simple RAG Systems: Minimal RAG application, built in Streamlit utilizing Qdrant as the vector store, Haystack for the orchestration pipeline.
  • Mixture of Experts classes: Laboratory classes about Mixture Of Experts architecture
  • Poker game: Online poker game built as a client-server architecture (Java)
  • CTF writeups: writeups for the CTF competitions I took part

Talks

My knowledge sharing activities: talks given on different conferences

Pinned Loading

  1. ChatterCast ChatterCast Public

    Podcast generation app using Chatterbox-Turbo TTS model

    Python

  2. efficient-attention efficient-attention Public

    Python

  3. fine-tuning fine-tuning Public

    Code for fine-tuning evaluation laboratory for NLP AGH classes.

    Jupyter Notebook

  4. frame-stacking-rl frame-stacking-rl Public

    Supplementary materials for my diploma project "Evaluating Frame Stacking Impact in Reinforcement Learning"

    Python

  5. GRPO GRPO Public

    Utilizing GRPO for fine-tuning Qwen3-0.6B model for recipes generation

    Python

  6. Ribbon-Exercise-Detection Ribbon-Exercise-Detection Public

    Jupyter Notebook