Kosovo Investment Data Platform
A public-data research scanner that turns fragmented Kosovo statistics into a cached, explainable decision-support interface.
The context
From an interesting idea to an inspectable system.
Official investment, tourism, housing, and business-registration data lives across separate statistical tables. The project brings those sources into one contract so regions and municipality-sector combinations can be compared without hiding the underlying numbers.
My contribution
What I built
- 01
Combined five ASKdata datasets with World Bank indicators through a reusable ingestion pipeline.
- 02
Normalized regional aliases and shaped the results into a single cached JSON contract.
- 03
Added retries, fallbacks, atomic writes, scoring, anomaly flags, forecasting helpers, and an interactive Streamlit interface.
- 04
Made the research-only limitation visible in the product rather than burying it in documentation.
Architecture
How the work moves
Evidence
Verified results
Honest evaluation
Limits and trade-offs
- The committed snapshot audited on 2026-09-04 was dated 2026-08-19; the app does not refresh automatically.
- Business registration data exposes a latest count and derived growth value, not a real multi-year series in the app contract.
- The business table still uses positional row selection and can change the selected municipality after a sector switch.
- The regional map needs live OpenStreetMap tiles even though the remaining application reads cached data.
Inspect the work
Stack and reproduction
- 1Clone the public repository and install app/requirements.txt.
- 2Run python pipeline/fetch.py to rebuild the current data snapshot, or use the committed snapshot.
- 3Run streamlit run app/app.py from the repository root.