Stress-Test Expert Analyst · Data Science & Quant Modelling @ the European Central Bank
Frankfurt am Main · MSc Mathematical Engineering (Quant Finance & AI) @ Politecnico di Milano
Hi, I’m Márton (Marci)—a data scientist, recovering competitive swimmer, and all-round ML tinkerer.
By day I design Monte-Carlo macro-financial simulations, AI strategy workshops, and stress-testing pipelines for Europe’s banking system. By night just swimming :)
Most of my professional work lives behind the firewalls of the European Central Bank and industry clients, so only a small slice ends up on GitHub. Still, feel free to browse the open repos below or reach out if something sparks your curiosity!
- Stress-Test Expert Analyst, Data Science & Quant Modelling — European Central Bank (Feb 2024 → now)
- 10 000-path Monte-Carlo stress scenarios, multi-agent RAG automation, and AI opportunity mapping across 109 banks.
- Part-time Data Scientist — Pannon Business Network
- Shipping Hungarian speech-to-speech assistants & computer-vision QC for smart-factory lines.
Auditable GNN Toxicity for Small Molecules
- Model: D-MPNN multi-task toxicity (Tox21) with Integrated Gradients → SMARTS mapping → toxicophore knowledge base for assay-aware, grounded explanations.
- Math bits: attribution reliability, threshold calibration, uncertainty awareness, and evaluation beyond AUROC (per-assay operating points).
- Why it matters: explanations chemists can interpret; artifacts that pass audit (versioned rules, provenance).
I have published the publish cleaned codes for reproducibility.
| Repo | Peek Inside |
|---|---|
| Auditable GNN for Molecules | xAI for Toxicity Prediction as a thesis for my masters. |
| ECBERT | Fine-tuning ModernBERT on monetary policy sentiment analysis. |
| Bayesian_Project | Spatio-temporal Bayesian modeling of ozone concentrations using STAN and hierarchical priors. |
| Non-parametric-Stats-Project | GAM, Mann-Whitney, Kruskal-Wallis & friends. |
| Applied-Statistics-Project | Functional data projections using GPs, LSTM, Prophet, KMeans, and polynomial/LOESS smoothing. |
| NLP-Project-mimic-ChatGPT-with-Orca | Fine-tuning Orca-Mini to imitate ChatGPT responses on a niche corpora. |
| NBA_shot_analysis | Tableau-based dashboard to explore how NBA scoring patterns shift by location and context. |
ℹ️ Some repos are private for IP or data-licence reasons. Ping me for a walkthrough if you’re evaluating my work.
Python · R · SQL · C++ · PyTorch · TensorFlow/Keras · scikit-learn · PyMC · AWS · GitLab CI/CD
- 🏊 Four-time World Junior swimming medallist turned quant nerd.
- 🌍 Native Hungarian, working in English, learning German & Italian.
- 🏔️ Currently hacking together an LLM-powered alpine-hut recommender.