3rd place
Open Data Heroes Hackathon with Demagog.sk AI Fact-Checker.
Source: official hackathon result page.
View hackathon coverageDeveloper and ML engineer from Bratislava
Building real systems from first principles.
I am a 16-year-old developer and ML engineer from Slovakia, currently in my second year at SPSE Halova. I also attend university-level lectures at FEI STU through the Habsburg program.
I focus on machine learning workflows, retrieval systems, and practical engineering over demo-only projects.
- 2nd-year student at SPSE Halova, Bratislava
- FEI STU lectures via the Habsburg program
- Primary language: Python, with C++ and SQL in active use
Placeholder
Credibility strip
Specific metrics with traceable sources.
~22,000
Political statements embedded and indexed in Qdrant.
Source: Qdrant collection count snapshot.
Qdrant photo pending24h
Prototype delivery window for the Pomelo Team hackathon build.
Source: hackathon format and project timeline.
View the websiteTwo-stage
LightGBM classifier and regressor architecture for rail delay research.
Source: project method documentation.
Research spotlight
ZSR Train Delay Prediction
Ongoing research paper
Active development
This is an ongoing ML research project, not a finished product launch. The goal is to model train delays with robust features and publish findings as a paper.
- Two-stage LightGBM setup: classifier for delay likelihood, regressor for delay magnitude.
- Graph-based features built with NetworkX to represent network context.
- Continuous MongoDB scraping pipeline for fresh operational data.
Current milestone: stabilize evaluation and prepare paper-ready experimental notes.
Built work
Shipped outputs and hackathon execution.
Awarded
Hackathon build
Demagog.sk AI Fact-Checker
3rd place project at Open Data Heroes Hackathon. Built semantic fact-checking flow using BGE-M3 embeddings, Qdrant vector search, RAG, and tool-calling for live web evidence.
- Indexed approximately 22k political statements for retrieval.
- Combined semantic search with generated verification flow.
- Designed for practical query speed and explainable output.
24h prototype
Pomelo Team Request System
Built in a 24-hour hackathon: a PHP, MySQL, and JavaScript request management system for an investor community with Gemini-powered request categorization.
- Structured request intake and tracking flow in one sprint.
- Added automated category suggestions to speed triage.
- Shipped a usable interface under hard time pressure.
In Development and Concept Work
In development
Slovak Political Fact-Checking Platform
Modular FastAPI monolith moving toward production readiness. Current focus is architecture hardening, reliability, and clean handoff between retrieval and verification services.
Technical stack
Tools I actively use.
Languages
- Python (primary)
- C++
- SQL
ML / AI
- LightGBM
- scikit-learn
- BGE-M3 embeddings
- Qdrant vector search
- RAG pipelines
- LLM tool-calling
Web
- FastAPI
- Vue.js
- TailwindCSS
- SQLModel
- PostgreSQL
- MongoDB
Infra
- Netlify
- Cloudflare
- Arch Linux / Hyprland
- Git
- QEMU / KVM
Education and trajectory
Growing from secondary school to research-level work.
Current
SPSE Halova, Bratislava
Second-year student building a strong software and engineering base while shipping projects outside the classroom.
University track
FEI STU lectures via Habsburg program
Attend selected university-level lectures in parallel, with a focus on deeper technical foundations and long-term research readiness.
Beyond code
Depth signals and personal interests.
I spend time on competitive programming, CTF and cybersecurity challenges, and deep ML internals. I wrote LightGBM pseudocode from scratch to understand the algorithm line by line. Outside engineering, I listen to power metal and care about sustainable systems design.
Competitive programming
CTF / cybersecurity
ML internals
Power metal
Sustainable systems
Contact