Data Analyst | ETL & Data Quality | SQL | Power BI | Business Intelligence

I build reproducible analytics projects that connect business questions with auditable data pipelines, reconciled metrics and decision-oriented dashboards. My portfolio focuses on evidence you can inspect: source code, tests, CI runs, data contracts, documented limitations and visible results.

I built a reproducible ETL and data-quality workflow to evaluate restaurant concepts and price positioning in Miami without publishing the private educational source.

• Stack: Python, pandas, Jupyter, data contracts, Ruff and GitHub Actions.
• Evidence: 24 automated tests, deterministic 750-customer synthetic demo and CI across Python 3.12–3.14.
• Decision: seafood and vegetarian concepts remain exploratory opportunities whose strength changes under conservative coverage thresholds.
• Reproduce it: run run_demo.bat from a clean clone.

Review the native figures PDF.

I turned six historical NBA datasets into an auditable ETL, a canonical SQL Server model and a six-page DirectQuery Power BI report.

• Stack: Python, pandas, SQL Server, Power BI, DirectQuery and GitHub Actions.
• Evidence: 161,111 input rows, 65,642 unique games, 15 Power BI objects, 13 tests and 89.92% core ETL coverage.
• Quality boundary: the repository processes 30.6 MB of versioned real inputs; it does not claim a completed 22 GB run.
• Result: English-authored six-page report, native high-resolution PDF, SQL reconciliation and a five-minute English reviewer guide. NBA is my main international case study.

Review the English report PDF.

I redesigned and hardened an executive Power BI case around revenue, margin, costs, prior-year performance and a correctly scoped USA drill-down.

• Stack: Power BI, DAX, Power Query, SQL and pbi-tools.
• Evidence: 60,398 fact rows, 35 normalized measures and 68 contexts reconciled through independent DAX and SQL paths.
• Result: four redesigned pages, versionable report source, PBIX/PBIT, automated validation and a published evidence pack.

Review the native report PDF.

I maintain a local-first Django application for financed sales, loans, installments, collections, backups and reporting. It demonstrates product delivery and provides a privacy-aware bridge into analytics.

• Stack: Python, Django, SQLite/PostgreSQL, Docker and pytest.
• Evidence: 337 functional tests passed in the current local check; the published release records 88% coverage and Windows migration/package checks.
• Analytics bridge: reconciled dashboard, an 11-file pseudonymized BI export, current demo screenshots and native/vector analytical PDFs.
• Distribution: v1.8.0 includes the full portable ZIP, update package, SHA-256 checksums and release audits.

Review the reconciled analytics PDF.

For a short overview across projects, see Questions, metrics and evidence. This is an editorial portfolio document, not a screenshot or an employment-history claim.

• Data: Python, pandas, NumPy, SQL Server, SQLite and PostgreSQL.
• BI: Power BI, DAX, Power Query, DirectQuery and Excel.
• Quality: pytest, Ruff, data contracts, reconciliations and GitHub Actions.
• Delivery: Git, Docker, release manifests, checksums and reproducible demos.

My background includes technical support, hardware and software troubleshooting, sales and cost management. I use that experience to translate technical findings into clear business decisions and practical deliverables.

• Spanish — Native.
• English — C1 Advanced (EF SET 63).

• Decision-oriented analytics and reporting.
• Reproducible ETL and data-quality workflows.
• BI architecture, metric governance and performance.
• Expanding toward cohort analysis, experimentation and forecasting with backtesting.

• LinkedIn
• Email: percy.marzoratti@gmail.com

⭐ I value transparent assumptions, reproducible evidence and conclusions that stay within the limits of the data.