About me
Ambitious Machine Learning Engineer and Data Scientist} with a strong foundation in data science developed through almost 5 years of academic studies and 2 years of hands-on industry experience. Passionate about applying state-of-the-art Machine Learning and Deep Learning techniques to solve real-world problems. Skilled in working with Large Language Models (LLMs), Natural Language Processing (NLP), MLOps, and multimodal learning. Dedicated to continuous learning through self-study, professional courses, and active participation in hackathons. Motivated to contribute expertise and enthusiasm to forward-thinking projects that advance AI and data science innovation.
Languages: Python, Java, R, SQL, HTML, CSS, JavaScript (basics), Bash
Frameworks: Django, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, NLTK, Keras,
Tensorflow, PyTorch, transformers, MLflow, Airflow
Tools: Git, Github Actions, Jupyter Notebook, PyCharm, Visual Studio Code, Anaconda, Docker,
Linux
Projects
This project, part of the Deep Learning course at Warsaw University of Technology, investigated various diffusion models for image generation. We implemented and tested three models—DDPM, DDIM, and PNDM—using the LSUN Bedroom dataset to generate 128x128 pixel images. We trained these models with PyTorch and the Hugging Face Diffusers library, comparing their performance over different epochs. Evaluations included FID scores and visual interpolation assessments. Comprehensive and reproducible results are detailed in the repository's README.
The project, undertaken as part of the Deep Learning course at Warsaw University of Technology, aimed to develop custom RNN architectures and test Whisper and Audio Spectogram Transformer and evaluate their performance. Utilizing the Speech Commands dataset, all models were trained using PyTorch. Comprehensive results, documented in the repository's README file, are reproducible for further analysis.
The project, undertaken as part of the Deep Learning course at Warsaw University of Technology, aimed to develop custom CNN architectures and evaluate their performance against pretrained models. Utilizing the CINIC-10 dataset, all models were trained using PyTorch. Comprehensive results, documented in the repository's README file, are reproducible for further analysis.
The project, undertaken as part of the Advanced Machine Learning course at Warsaw University of Technology, aimed to implement Logistic Regression from scratch with 3 different optimizers: SGD, IRLS and Adam. Additionaly, evaluate their performance in binary classification task alongside with LDA, QDA, Decision Tree and Random Forest. Comprehensive results, documented in the repository's README file, are reproducible for further analysis.
A Django web app can help students manage their studies by providing a platform to create and manage tasks and projects. Each task can have a priority level, and the app also has a translator tool to help with notes and assignments in different languages, which can be useful for international students or those studying in a foreign language.
Dashboard created in Streamlit showing analysis of the data from Messenger. Project required pipeline for data processing and analysis.
Two projects from my studies at Warsaw University of Technology. The first one is a classification task, where I had to predict the activity of a person based on the data from their phone. The second one is a clustering task, where I had to cluster documents based on their content.
A poster on environmental protection within the subject of Data Visualization Techniques. The poster was created in Canva and contains a lot of charts and graphs created using R language.
Experience
July 2024 - Present: Machine Learning Intern - Snowflake
- Gaining in-depth knowledge of Snowflake's product offerings
- Developing chatbots for internal stakeholders using Retrieval-Augmented Generation (RAG)
- Deploying LLMs and VLMs on Snowpark Container Services (SPCS)
- Creating a multi-modal RAG chatbot capable of processing both images and text
- Exploring different quantization techniques and experimenting with deployment using llama.cpp, vLLM
- Fine-tuning Llama 3.1 model using Low-Rank Adaptation (LoRA)
- Utilizing technologies such as Python, Snowflake Cortex, LangChain, Docker, Streamlit, Trulens and vLLM for LLM serving and inference and implementing REST APIs and managing LLMOps
December 2023 - March 2024: Junior Machine Learning Engineer - Grid Dynamics
-
Utilizing acquired expertise in 2 client projects
-
NLP project
- Enhancing product recognition algorithms by refining pattern matching using Regex, and incorporating algorithms based on Levenstein distance and fuzzy search.
- Participating in a dynamic three-person team, actively contributing to collaborative problem-solving initiatives aimed at addressing challenges and optimizing product recognition algorithms.
- Data cleaning and preprocessing using pandas
-
Optimization project
- Assisting in refactoring the code base and addressing performance issues on Databricks
-
NLP project
June 2023 - December 2023: Machine Learning Engineer Intern - Grid Dynamics
- Acquiring Proficiency in the Following Areas:
- Python
- SQL
- Flask and FastAPI
- Docker conterization
- Apache Airflow
- MLflow
- PySpark Data Processing
- Cloud Expertise (AWS, GCP)
- Jenkins & Terraform
- Achieved certification:
July 2022 - December 2022: NLP Intern - Samsung R&D
- Development of Bixby, working on automation tools for linguist
- Improving deep learning model performance in NLU area (mainly Tensorflow/transformers, significant improvement 5% accuracy on production data)
- Implementing features in Android project written in Java
- Technical Skills: Python with Tensorflow and HuggingFace, Java (Android project), Linux tools, Scripting (Bash), Git, Github, Github Actions.
- Soft Skills: Teamwork, Time Management, Communication.