Особисті дані приховані

PROFILE
Roman Trehub Python developer focused on building AI-driven systems and
machine learning solutions. Experienced in data collection,
Data Scientist preprocessing, and analysis of real-world datasets, as well as
deep learning, natural language processing, and computer vision.
CONTACT Strong interest in generative AI, LLM-based systems, cloud
technologies, and building scalable, data-driven applications.
Email
[відкрити контакти](див. вище в блоці «контактна інформація») EDUCATION
Phone Computer Science - Master of Science 2025 - Present
[відкрити контакти](див. вище в блоці «контактна інформація») (MSc)
LinkedIn Technical University of Kosice, Faculty of Electrical
Engineering and Informatics
[відкрити контакти](див. вище в блоці «контактна інформація»)
Currently pursuing MSc in Intelligent Systems with focus on machine
GitHub learning and data-driven methods. Includes NLP, computer vision,
github.com/Romaha095 predictive modeling, and development of scalable AI-powered
systems with real-world applications.
TECHNOLOGIES
Computer Science - Bachelor of Science (BSc) 2022 - 2025
Programming: Python, C/C#
Libraries: Flask, FastAPI, Pandas, Technical University of Kosice, Faculty of Electrical
Engineering and Informatics
NumPy, SciPy, Matplotlib, Seaborn,
OpenCV, tqdm Strong foundation in algorithms, data structures, and software
Machine Learning Libraries: PyTorch, engineering, with focus on data processing, applied machine
Scikit-learn, Transformers, Hugging learning, and building simple ML pipelines. Worked with databases,
Face, NLTK, PEFT, vLLM, TGI, systems, and networks while developing practical AI-oriented
LangChain solutions and robotics-related projects.
Databases: PostgreSQL, MongoDB
Cloud and MLOps: AWS, Azure,
Docker, CI/CD, Model Deployment
PROJECTS
Tools: Git, Linux, Jupyter Notebook Search for Mayan Glyph Signs in Dictionaries
Methodologies: Agile, Scrum Bachelor's Thesis
LANGUAGES Designed and implemented an end-to-end pipeline to segment
Mayan glyphs from line-drawing transcriptions and match them
Slovak - C1 against structured glyph dictionaries.
English - B1 Trained and evaluated a glyph embedding model for
similarity-based retrieval, enabling automated matching as part of
Russian - Native
a larger Mayan text translation workflow.
Ukrainian - Native
Tech: PyTorch, Transformers,OpenCV, NumPy, pandas

Real-time Parking Availability Mobile App
Team project with mentors from Fpt Slovakia
Developed backend services for a mobile app displaying available
parking spots with filtering and navigation.
Processed real-time sensor and camera-based data for occupancy
detection and updates.
Tech: Python, Flask, AWS, Docker

RAG Document Assistant Chatbot
University Project
Developed an end-to-end RAG pipeline for document question
answering.
Built ingestion pipeline including PDF parsing, hierarchical
chunking and embeddings.
Implemented hybrid search, reranking and web chat interface with
citations.
Tech: Python, LangChain, Hugging Face, Transformers, Ollama,
FAISS, PEFT, Docker

Roman Trehub | Data Scientist