My latest work

DuckHaven is a sovereign analytics platform for small teams, powered by DuckDB and Iceberg. DuckHaven is a personal project that I started to see whether open source table formats, query engines and catalogs can replace the need for a cloud data warehouse.

A lightweight, self-hosted container vulnerability scanner with a hub-agent model, powered by Trivy. Built for homelabs and small multi-server Docker environments.

A Python app that streams real-time Formula 1 timing from the OpenF1 API and visualizes lap times, pit stops, and session statistics as a race unfolds, with customizable filters to slice the data.

A CLI that crawls a news outlet's sitemap for a given day, then runs Hugging Face summarization and sentiment models over each article — quick reads plus a bias score — and lands the results as Parquet on S3.

This CLI allows you to fetch all the developer diaries of a PDX game to a specified location (local or s3 bucket). It's aim is to help fellow PDX gamers to analyze diaries provided by the development team.

Scraping data from different sources often leads to discrepancies, like missing information. This project aims to solve this issue in the supermarket groceries domain by deploying NLP techniques to classify products into distinct categories.

This was my first project working with geojson data. I used geopandas coupled with Kepler-GL to visualize the public transport system in Budapest, Hungary.

Built an econometric model revolving around rental apartments in Budapest. This was the first time I experimented with regularization (lasso, ridge, elasticnet).