Predicting sovereign rating changes 30 to 90 days ahead from daily market data with an LSTM and MLP fusion model.
Star ratings and review text disagree in 44.5% of 100,000 Amazon reviews. What drives the gap?
A price tracker for used PC parts that crawls three Hungarian marketplaces, keeps the full history of every listing and turns messy titles into structured products.
A hidden-movement chase game on the real Budapest timetable, balanced with thousands of bot-vs-bot simulations.
XGBoost and LSTM trend models, a score-based ETF allocation and a news sentiment sleeve, running on paper money with a web dashboard.
AWS Lambda pulls weather data into an S3 data lake, Glue and Athena join it with inverter telemetry to measure heat-related energy loss.
CSV to Parquet conversion, Glue crawlers, Athena views and analysis on 70 years of global fisheries data.
I study Data Science in Business at Corvinus University of Budapest (2023 to 2027). My thesis uses large language models to extract structured knowledge from text. I like projects where data meets a real system: collecting it, modelling it and building the tool people actually use.
I worked as a controlling intern at Dometic, and I build websites and AI automations for small businesses on the side. Outside of code I am a European Champion sabre fencer.