# Roman Žydyk

Czech version: https://romanzdk.com/cs/index.md

Data platforms. Reliable Python. Practical AI.

I work across data engineering, software quality, and practical automation. I build tools, connect systems, and turn complex technical topics into useful learning material. I’m interested in AI agents where they help get real work done — with clear limits and results that can be checked.

- Email: romanzdk@gmail.com
- LinkedIn: https://www.linkedin.com/in/romanzdk/
- GitHub: https://github.com/romanzdk
- Website: https://romanzdk.com/
- AI & Automation: https://romanzdk.com/ai/index.md

## Background & Approach

My background spans business intelligence, analytics, and data engineering, including work at PwC and Quantlane. My experience includes data platform architecture, Python pipelines, software quality, and operational reliability.

As a technical lead, I’ve shaped the team’s technical direction and architecture decisions, held regular 1:1s, and helped engineers grow their technical skills.

I value maintainable systems, explicit trade-offs, and practical solutions over unnecessary complexity.

Areas:

- Data engineering & platforms: SQL, Airflow, pandas, Polars, Arrow/Parquet, Apache Iceberg, PostgreSQL, TimescaleDB, Redis, RabbitMQ
- Python & software quality: FastAPI, asyncio, typing, testing, linting, packaging, code review
- Cloud & operations: Kubernetes, Docker, Argo CD, GitLab CI, AWS, Azure, Prometheus, Grafana, Sentry, Kibana
- AI agents & automation: coding agents, agent review workflows, home automation
- Technical leadership & teaching: technical direction, architecture decisions, 1:1s and mentoring, hiring and onboarding, courses and workshops

Professional background on LinkedIn: https://www.linkedin.com/in/romanzdk/

## Selected work

### Live Czech election results (2021, 2023, 2025)

- Problem: clean, fast election-night results without ads and commentary.
- Role: built it alone – data loading, backend, frontend, deployment.
- Decisions: the first version was a Power BI report refreshed by hand. Later versions used a Python backend that polled the official results feed of the Czech Statistical Office, stored it in PostgreSQL and showed it in a Streamlit app, each part in Docker. In 2025 the backend was hand-written and the frontend generated by an AI coding agent.
- Code: the 2023 version is public – backend https://github.com/romanzdk/volby-2023-backend and frontend https://github.com/romanzdk/volby-2023.

### Other projects

- [S3 web browser](https://github.com/romanzdk/s3-web-browser): built a simple web UI for browsing S3 storage and sharing files through time-limited links. Related: [S3 Commander](https://github.com/romanzdk/s3-commander), a Total Commander-style, two-pane S3 browser with server-side move and copy (FastAPI).
- [Deadly car accidents in Czechia](https://github.com/romanzdk/DSBI-cz-car-accidents): winning team project at the Data Science & BI Academy (VŠE, 2019/2020). Merged 12 years of Czech police accident data and compared classifiers; a collision with a pedestrian was the strongest predictor of a deadly accident.
- [Tableau Public dashboards](https://public.tableau.com/app/profile/roman.zydyk): 13 public dashboards, from a decade of Forbes billionaires (with scraped data) to table tennis in Czechia and an equipment selector.

### Česko.Digital volunteering

Volunteered with [Česko.Digital](https://www.cesko.digital/) as a data engineer and consultant on several projects. Mostly projects that helped people after the start of the war in Ukraine, such as [Movapp](https://www.movapp.cz/), [Stojíme za Ukrajinou](https://stojimezaukrajinou.cz/), UMapa and Místo ve škole, automating and connecting data loads. Also internal reporting pipelines in Keboola, SQL, Python and Snowflake.

### Open-source contributions

Contributed fixes and features to open-source projects, including [exchange_calendars](https://github.com/gerrymanoim/exchange_calendars), the official [Massive (Polygon) Python client](https://github.com/massive-com/client-python), [bellybutton](https://github.com/hchasestevens/bellybutton), [youtrack-sdk](https://github.com/moneymeets/youtrack-sdk) and Česko.Digital’s [Movapp](https://github.com/cesko-digital/movapp).

## AI & Automation

- AI-assisted development: using coding agents for implementation, testing, and review, with an emphasis on checking the result rather than trusting the output.
- Assistants, memory & integrations: exploring assistants that connect tools, retain useful context, and help with research and recurring tasks.
- Practical automation: connecting APIs, browser workflows, and voice interfaces — while keeping simple or critical operations deterministic.

Approach and examples: https://romanzdk.com/ai/index.md

## Courses & Teaching

- [Database & SQL](https://turinginstitut.cz/kurzy/relacni-databaze-a-sql/) (live online workshops, Czech): relational databases and SQL, taught live for Turing Institut (formerly BeeIT).
- [Python: Code quality](https://skillmea.cz/online-kurzy/python-code-quality) (online course, Czech): writing Python that is typed, tested, documented, secure and fast enough.
- [Azure cloud basics](https://skillmea.cz/online-kurzy/azure-cloud-zaklady) (online course, Czech): cloud concepts and a practical tour of Azure services.
- [Tableau in practice](https://skillmea.cz/online-kurzy/tableau-prakticky) (online course, Czech): using Tableau for data visualization in everyday work.

## About

I started in BI and reporting, moved through data science, and ended up in data engineering. Outside work, I’m into table tennis, cars, travel, and building tools that make everyday life easier.

MSc (Ing.) in Information Systems and Technologies – Business Intelligence, Prague University of Economics and Business, 2022, with a thesis on MLOps with Kubeflow.

## Contact

Interested in data engineering, Python, or practical AI? Get in touch: romanzdk@gmail.com