Controlling Intern, Dometic
August 2025
- Automating Excel-based workflows
Profile
Data Science in Business BSc student at Corvinus University of Budapest. I am interested in Python, machine learning and large language models, and work with them in personal projects and in my thesis: I built a data collection system with a locally hosted language model, and my thesis extracts structured knowledge from text. I gained corporate experience as a controlling intern.
Experience
Web Development and AI Solutions, self-employed
- Lead finder tool using Google Places data with prospect scoring
- Live Google review request automation for a Budapest flower shop (WooCommerce webhook, dedicated sending domain with SPF/DKIM/DMARC)
- AI chatbots and automations, plus websites and webshops (WordPress, WooCommerce) for small and medium-sized businesses
Projects
Credit Risk Early Warning System: course project, deep learning part
Python, PyTorch, pandas, scikit-learn
Late-fusion model of an LSTM and an MLP encoder over daily market data (CDS, yields, FX volatility) and monthly macro data, predicting S&P sovereign rating changes 30-90 days ahead. Chronological validation and class weighting for rare events.
The Sentiment Gap: text analytics team project
Python, NLTK, scikit-learn, SciPy
On 100,000 Amazon reviews, star rating and text sentiment disagree in 44.5% of cases (chi-square, Cramér's V = 0.32). VADER, TF-IDF with logistic regression (macro F1 0.655), LDA topic modelling.
apro-tracker: data collection and price tracking for used hardware listings
Python, Scrapling, Ollama, SQLAlchemy, Docker
Collects listings from three Hungarian second-hand marketplaces, keeps full price history, and turns messy titles into structured products with a rule parser and a locally hosted language model.
Fare Dodger: simulation and game on Budapest public transport
Python, BKK GTFS, self-written RAPTOR router
Bot-vs-bot simulations (200 games per matchup, Wilson intervals) to balance the game rules on the real BKK timetable.